Astm stp 266 1960

When studying the reactions under- gone by 3CaO.AI~O3, it must be remem- bered that this compound can react with water to form hydrated compounds and with calcium sulfate, calcium hydrox

Reg U.S Pat Off

A S T M Special Technical Publication No z66

Price 87.50; to Members $6.00

Published by the AMERICAN SOCIETY FOR TESTING MATERIALS

I916 Race St., Philadelphia 3, Pa

©BY A~R~c~,~ SOCIETY FOR TEsT~O MATEm~a.S 1960

Library of Congress Catalog Card Number: 60-9520

Printed in Baltimore, Md

June 196o

F O R E W O R D

The use of water-reducing admixtures and set retarders in concrete has grown continuously since their introduction over 25 years ago, with a present estimated usage in the production of 25 million cubic yards of concrete annually in the United States alone Further indication of current interest

is evidenced by the fact that ASTM Committee C-9 on Concrete and Con- crete Aggregates and certain public agencies are actively drafting standard methods of test and specifications to govern the purchase and performance requirements of these admixtures for concrete

The future of better concrete lies in an increased understanding of con- creting materials and the best manner of combining them to produce the maximum in strength and durability It is with this in mind that Committee C-1 on Cement and Committee C-9 have jointly sponsored this Symposium, held during the Third Pacific Area National Meeting of the Society from October 11-16, 1959, in San Francisco, Calif

The Symposium consisted of ten papers and a summary It is of interest to note that four of the papers represent the joint contribution of four principal producers of admixtures The remaining papers were prepared by repre- sentatives of consumer interests, research organizations~ and the cement industry

The Symposium was held during two sessions on Wednesday, October 14 Professors R E Davis and Milos Polivka, both of the University of Cali- fornia, presided over the two sessions respectively

The Joint Symposium Committee, representing Committees C-1 and C-9, included the following members:

R L BLAINE, National Bureau of Standards

JosEPH E G~Y, National Crushed Stone Assn

E C HmGINSON, U S Bureau of Reclamation

THOMAS B KENNEDY, Corps of Engineers

WnJ.IAM LERCH, Portland Cement Assn

W J McCoY, Lehigh Portland Cement Co

RICg_AI~ C MI~LENZ, Master Builders Co

NoTE. The Society is not responsible, as a body, for the statements

and opinions advanced in this publication

C O N T E N T S

P A G E

Introduction Bruce Foster 1 Actions of Calcium Sulfate and Admixtures in Portland Cement Pastes W C Hansen 3 Discussion 25 Structural and Lean Mass Concrete as Affected by Water-Reducing, Set-Retarding Agents George B Wallace and EIwood L Ore 38 Discussion 94 Observations in Testing and Use of Water-Reducing Retarders Lewis H Tuthill, Robert F Adams, and John M Heroine, Jr 97 Discussion 118 Effect of Water-Reducing Admixtures and Set-Retarding Admixtures as Influenced

by Portland Cement Composition Milos Polivka and Alexander Klein 124 Field Experience Using Water-Reducers in Ready-Mixed Concrete E L Howard,

K K Griflfiths, and W E Moulton 140 Discussion 148 Detection of Lignosulfonate Retarder in Cement Suspensions and Pastes E G Swensen and T Thorvaldson 159 Discussion 169

Introduction to Producers' Papers on Water-Reducing Admixtures and Set-Retard- ing Admixtures for Concrete M E Prior and A B Adams 170 Effect of Water-Reducing Admixtures and Set-Retarding Admixtures on the Prop- erties of Plastic Concrete C A Vollick 180 The Effect of Water-Reducing Admixtures and Set-Retarding Admixtures on the Properties of Hardened Concrete D R MacPherson and H C Fischer 201 Water-Reducing Admixtures and Set-Retarding Admixtures for Concrete: Uses; Specifications; Research Objectives Richard C Mielenz 218 Discussion 234 Summary Bruce Foster 240

SYMPOSIUM ON E F F E C T OF W A T E R - R E D U C I N G A D M I X T U R E S AND S E T - R E T A R D I N G A D M I X T U R E S ON P R O P E R T I E S

OF C O N C R E T E

I N T R O D U C T I O N

BY BRUCE FOSTER 1

Admixtures for portland cement con-

crete are those ingredients which are

added to the primary constituents (port-

land cement, aggregates, and water) to

(1) improve or modify the properties of

the concrete, (2) compensate for some

deficiency in a primary constituent, or

(3) effect a reduction in cost Some ma-

terials which we class as admixtures,

such as pozzolan and blood, were con-

crete ingredients used by the Romans

Because the addition of another in-

gredient is very likely to require addi-

tional control and technical skill on the

part of the concrete producer and in-

spection agency, and extra facilities for

handling and proportioning, and because

admixtures in general were originally

frowned upon by the cement and con-

crete industries, these materials were for

many years slow in gaining general ac-

ceptance However, the recognition of

the role of entrained air in imparting

frost resistance to concrete has led to al-

most universal acceptance of the value of

air-entraining admixtures and to de-

parture from the previously widely-held

concept that all admixtures were of

doubtful value

Appreciation of the advantages in the

use of another group of admixtures, the

pozzolans, was evidenced by the "Sym-

1 National Bureau of Standards, Washington,

1954, for natural pozzolans in 1957, and which is now preparing specifications for accelerators, for set retarders, and for water-reducing admixtures

The last two mentioned, water-reduc- ing and set-retarding admixtures, are the subject of this symposium As its name implies, a water-reducing admixture, when added to a concrete, permits a re- duction in mixing water with no loss in slump, or, if the water content is main- tained constant, produces an increase in slump The name is not wholly descrip- tive, however, because as will be brought out during the symposium, the benfit in strength at constant slump are normally greater than would be expected from the resulting reduction in water-cement ratio

A set-retarding admixture reduces the

2 Am Soc Testing Mats (1950) (Published

as separate publication A S T M 8 T P No 99.)

2 S ~ P O S I ~ ON ADMIXTURES IN CONCRETE

early rate of hardening and so permits

the concrete to be handled and vibrated

for an additional period after mixing

The principal agents now in use fall in

one of two classes: (1) lignosulfonic acids

and their salts and (2) hydroxylated

carboxylic acids and their salts Both

classes of materials, when added to con-

crete, reduce the water requirement and

also retard the set Modifications and de-

rivatives of these materials may retain

the water-reducing property of the ad-

mixture without modifying the harden-

ing rate, or may even accelerate the set

The Symposium Committee repre-

senting Committees C-1 on Cement and

C-9 on Concrete and Concrete Aggre-

gates, planned to bring together informa-

tion on admixtures as follows: (1) the

mechanisms by which these materials

modify concrete properties; (2) the ef-

fects of the admixtures on the properties

of plastic and hardened concrete and the

variation of these effects depending upon

the other materials involved, the type of

concrete, and the existing temperature; (3) the types of construction and the conditions under which their use is par- ticularly advantageous; (4) the problems

in control and application brought about

by their use; (5) the problems of prepar- ing adequate purchase specifications; and, (6) research under way to produce even better and more reliable admixtures and thereby better and more economical con- cretes

To accomplish this, contributions were sought from a variety of sources includ- ing universities and government agencies,

as well as the producers of portland ce- ment, ready-mixed concrete, and ad- mixtures A great amount of laboratory and field data are presented as well as descriptions of field experience with the use of the various products in a wide range of applications Of necessity, there

is some overlapping in the treatment of the subject matter, but this will be found

to have more advantages than disad- vantages

ACTIONS OF CALCIUM SULFATE AND A D M I X T U R E S

in the surfaces of the crystals cf the cement minerals and then diffuse into the crystals to produce the reaction products responsible for the setting and hardening of the cement paste The abilities of organic compounds to retard the rates of these reactions, cause dispersion of the particles, reduce water require- merits, and act as grinding aids are explained in terms of adsorbtion of mole- cules and negative ions

The purpose of this paper is to suggest

a mechanism whereby admixtures affect

certain properties of pastes of portland

cement and water Very early in the de-

velopment of the cement industry (1), 2 it

was found necessary to use calcium sul-

fate with most clinkers to produce ce-

ments with satisfactory setting charac-

teristics

A given admixture may show different

behaviors with ~different clinkers, but it

is likely to show a consistent behavior if

the clinkers areground with calcium sul-

fate as is done in making portland ce-

ment This paper is, therefore, concerned

with the mechanism whereby calcium

sulfate influences the reactions of the

cement minerals With water and the

i Director, Research Laboratories, Universal

A t l a s C e m e n t Division of U n i t e d States Steel

Corp., Burlington, Ind

* T h e boldface n u m b e r s in p a r e n t h e s e s refer

to t h e list of references a p p e n d e d to t h i s paper

mechanism whereby admixtures modify these reactions and alter the physical properties of the cement I t makes no

a t t e m p [ to deal with reactions that might occur with the portland cement minerals in the absence of calcium sul- fate

Rankin and Wright (2) showed that, if

a portland-cement clinker consisted only

of CaO, A12Os, and SiO2, it would, at equilibrium, consist of the three com- pounds: 3CaO.SiO~, 2CaO.SiO2, and 3CaO-Al203 Since that time, many investigators have added information on the roles of the minor oxides (Fe203, MgO, N a 2 0 , and K~O) in the constitu- tion of portland-cement clinkers The various phases of the constitution of these clinkers were reviewed in 1952 by Jeffery, Nurse, Ordway, Malquori, Ci- rilli, Newkirk, Insley, and by a number

of discussors of their papers (3)

4 SYMPOSIUM ON ADMIXTURES IN CONCRETE

Those reviews show that the principal

compound of such clinkers, which is

3CaO.SiO~ in the CaO-AIK)3-SiO2 sys-

tem, probably contains some MgO and

Al203 and may have the composition

54CAO 16SIO2- MgO Al~O3 • They show

that the 2CaO.SiO~ phase of the CaO-

Al~Os-SiO~ system may be a solid solu-

tion of 2CaO-SiO2-23CaO.KzO 12SiO2

in portland-cement clinkers Also, that

the Al~O3 may exist as 3CaO.Al2Os,

8CaO-Na20.Al203 and as a solid solu-

tion of 2CaO Fe20~ and 6CaO 2Al20,

Fe,Os The alkalies K~O and NaK)

combine with SO3 in the clinkers to form

alkali sulfates, and those in excess of the

SO, form the alkali-bearing phases

listed above

EARLY STUDIES WITH CALCIUM SULFATE

Bates and Klein (4) and Bates (5)

studied several of the pure compounds

of the CaO-AI~O~-SiO~ system and found

that:

1 A paste of powdered 3CaO-AhOa

and water steamed and hardened very

quickly;

2 The rate of this reaction could be

retarded by blending the powdered

3CaO.Al~O8 with calcium sulfate and

that small amounts of calcium hydroxide

added to this blend increased the re-

tarding power of the calcium sulfate;

3 3CaO-SiO~ augmented the retard-

ing power of calcium sulfate, probably

because of the Ca(OH)2 released by the

hydrolysis of this silicate;

4 3CaO.SiO~ reacted fairly rapidly

with water and that the rate of this re-

action was accelerated by calcium sul-

fate; and

5 3CaO-SiO~ and the other com-

pounds in portland-cement clinkers,

when mixed with water, did not require

any retarders for satisfactory placing in

molds

Phillips (~), working with Bates and

Klein, studied the reactions of 3CaO- AltOs and water from the standpoint of colloidal chemistry He pointed out, as earlier investigators had, that 3CaO Al203 and the calcium silicates hydro- lyzed in water with the splitting off of Ca(OH)~ and the formation of less basic aluminates and silicates I t appears from his paper that he had reached the con- clusion that 3CaO Al~O~ in a paste with water did not dissolve but reacted with water to split off Ca(OH)~ and form a dispersion of solid calcium aluminate

He states:

"In the normal pastes, we should suppose that the very fine material would be hy- drated and dispersed to a sol which would later coagulate to form a gel with elimina- tion of water, while the coarse material would very slowly hydrate and distend to form a g e l The control of hydration is af- fected to a considerable extent by the ad- sorption of electrolytic ions In examining the aluminate we find a good example of adsorption, for the treatment with water changes the composition of the outer surface only, a hydrated film being formed while the interior is unchanged Any ion activity will therefore be confined to the surface It has been previously stated that lime water effects a greater dispersion of the aluminate than does pure water This is due to the fact that the positive calcium-ions of the lime hydrate are adsorbed by the hydrated aluminate When the concentration of the adsorbed ions becomes sufficiently great, or when a sufficient number of ions have been adsorbed, their charges repulse each other and the aluminate is dispersed In order to get the repelling action we must suppose that the positive calcium ions are adsorbed more than the negative hydroxyl ions The results show that this occurs or, in other words, there is a preferential adsorption of one kind of ion."

Roller if), from his own work and that

of others, concluded that calcium sulfate retarded the setting of portland cement

by a combination of two reactions First, it reacted with alkali hydroxides

HANSEN ON ACTIONS OF CALCIUM SULFATE 5

released from the cement to form

Ca(OH)~ as illustrated in Reaction 1,

where M is either or both K and Na

Then, the Ca(OH)~ reacted with 3CaO

A1203 as illustrated in Reaction 2

CaSO4 + 2MOH M2SO4 + Ca(OH)v (1)

3CaO-Al2Os + Ca(OH)2 + I2Hg:)

= 3CaO.Al203-Ca~OH)v 12Hg) (2)

I t is well known (8) that 3CaO.A12Oa

Ca(OH)~ as shown in Reaction 3 and

with CaSO4 in a solution saturated with

Ca(OH)2 as shown in Reactions 4 and 5:

3CaO- AI~O, -I- 6HK) 3CaO Al2Ov 6Hd3 (3)

3CaO.Al~O, + CaS04 -~- 13Hd3

= 3CaO Al,~Ov CaSO4-13Hd3 (4)

3CaO.Al~Oa + 3CaSO4 + 32H#3

ffi 3CaO • Al,~h- 3CaSO,v 32H~,O (5)

Since the publication by Roller, it has

been established (9) that 3CaO.A1203

CaSO4.13Hg) form a complete series of

solid solutions and that 3CaO.Alg)3

3CaSO¢32H20 also form a complete

series of solid solutions These series

might be represented as follows:

3CaO.AlsO3 + XCaSO4 + YCa(OH)~ + ZH~O

= 3CaO AI20, XCaSO4

• YCa(OH) 2" ZH20 (6)

where X m a y vary between 0 and 1 and

Y m a y v a r y between 0 and 1 for the

first series, and where X m a y v a r y be-

tween 0 and 3 and Y m a y v a r y between

0 and 3 for the second series

Roller believed that Reactions 3, 4,

and 5 would cause flash setting and that

this could be prevented b y forming a cer-

tain amount of 3CaO-Al~Oa- Ca(OH)~

12H~O as illustrated in Reaction 2 on

the surfaces of the grains of 3CaO A1203

Also, that Reactions 2 to 6, inclusive,

could take place without either the 3CaO.AI~O3 or the reaction products going into solution I n other words, these could be considered to be solid state reactions in which one of the re- actants reacts as a solid with other re- actants which are in solution to form directly solid reaction products This conclusion was based on the observation

t h a t extracts from cement pastes contain only very minor amounts of A1208 and SiO2

I t is indicated in Reaction 1 that the alkalies entered the liquid phase of ce- ment pastes as hydroxides I t is now known (10) that most of the alkalies that enter the liquid phase during the first few minutes do so as alkali sulfates However, since 3CaO.Alg)3 can react very rapidly with CaSO4, the small amount of SOs that enters as an alkali sulfate is converted to CaSO4 b y reaction with Ca(OH)~ released from other ce- ment minerals and is removed from the solution very quickly This leaves the alkalies in the solution, in the absence of added salts, as the hydroxides and Rol- ler's Reaction 1 is essentially correct when it is recognized that the hydroxide entered the solution principally as Ca(OH)2

Roller believed that work of others had shown that calcium sulfate did not retard the rate of reaction of 3CaO Ald33 with water However, Bates (sO, working with pure 3 C a O A l g ) , , states:

"Five per cent of plaster of Paris retarded the steaming sufficiently to permit the mak- ing of test pieces Five per cent of plaster in the presence of two per cent of hydrated lime produced a much more marked retarda- tion of the initial set but was accompanied

as before by an evolution of steam Increas- ing amounts of plaster produced increased retardation."

In analyzing the results of tests made

on pastes in which all reactants, except water, are initially solids, it must be

6 SYm~osnm oN ADMIXruKES IN CONCRETE

recognized t h a t compounds such as

3CaO.AI~Os, 3CaO.SiO~, and plaster

of Paris are unstable in water whereas

CaSO~-2Hg3 and Ca(OH)2 are stable

Hence, even if the ions of the last two

are capable of retarding the rate of re-

action of 3CaO.AI~O3 with water, the

3CaO.A16)3 in a paste made from the

three solids m a y have undergone con-

siderable reaction before the other two

had dissolved sufficiently to exert a re-

tarding effect However, when plaster of

'Paris is used, it m a y dissolve fast enough

to be fairly effective as a retarder I t

seems that the experiments b y Bates

show that both calcium sulfate and cal-

cium hydroxide can retard to some extent

the rate at which 3CaO.AI~O, reacts

with water and that calcium hydroxide

m a y be more effective than calcium sul-

fate

Forsen (11), from a review of the liter-

ature and from extensive studies of his

own, makes the following statements:

"When pordand cement reacts with

water, the components enter into solution

in the same stoichiometric proportions as the

water-free components have, forming solu-

tions which are supersaturated with respect

to hydrates as solids."

This behavior of the cement minerals

probably was first postulated b y Le

Chatelier (12)

This conclusion b y Forsen was based

on experiments in which 0.2 g of a com-

pound was treated with a liter of water

He states, however, that Flint and Wells

(34) "examined the behavior of 2CaO

SiO2 and 3CaO-SiO2 in water (20 g per

liter) and found that, even after short

periods of shaking, the proportion of

lime to silicic acid was higher in the solu-

tion than in the water-free components."

I t seems very unlikely that any of the

calcium silicates and aluminates of ce-

ment clinkers dissolve in pastes as For-

sen states

"Most technical cements containing alkali are quick setting without gypsum Alkali aluminates and silicates enter instantane- ously into solution and coagulation between silicate and aluminate occurs The quick set

of such cement cannot be prevented by the addition of lime, because the solubility of the lime is low in the instantaneously formed alkali hydroxide solution and because the lime is dissolved touch,lower than the alkali aluminate and the alkali silicate The addi- tion of readily soluble gypsum or other calcium salts neutralizes the alkali hydroxide because calcium hydroxide is less soluble than alkali hydroxide A saturated or super- saturation of cateium hydroxide in a calcium and alkali salt solution is formed and the aluminates are immediately precipitated A protecting film is formed around the cement grains, which retards the dissolution of the

pounds."

I t is now well known that, generally,

a relatively large part of the alkalies in cement occur as either K~SO4 or Na~SO4

or as a solid solution of the two (New- kirk (3)) and that it is these alkali salts that are dissolved almost immediately when the cement is mixed with water (13) Within a minute after such clinkers are mixed with water, the liquid phase contains relatively high concentrations

of alkalies, lime, and sulfate but only traces of alumina and silica

Schlapper (14), in discussing the retard- ing effect of calcium sulfate states:

"According to Forsen, this is supposed to prevent the precipitation of silicic acid by the aluminates, which in turn is supposed to retard the setting process We were able to confirm by our experiments that silicic acid and alumina precipitate each other floccu- larly, but, on the other hand, we also estab- lished in accordance with the investigations

on equilibrium mentioned above that the quantities of silicic acid in solution are very slight; the question, therefore, arises whether the quick setting of a mixture of 3CaO-SiO~ and aluminate is caused chiefly by this de- flocculation."

H A N S E N ON ACTIONS o~" CALCIUM SULFATE 7 Forsen, in reply to a discussion of his

paper by Roller, states: " I have also

verified analytically and synthetically

- - t h e formation of a protective film of

tetracalcium aluminate when anhydrous

calcium aluminates react with limewater

This fact was first established micro-

s c o p i c a l l y - b y Assarson."

Strelkov (15), in reviewing the work

carried out by Baikov during the years

1923 to 1932, states: "Baikov demon-

strated that the colloidal stage is com-

mon to all cementing materials As an

intermediate stage, it brings about the

recrystallization of the initially formed

hydrates due to the chemical reactions

of cement particles with water without

their preliminary dissolution."

In this theory, Strelkov believes that

both plaster of Paris and portland cement

react with water without going into

solution to form products with high

solubilities around the anhydrous grains

These new products then dissolve and

which the stable reaction products pre-

cipitate He points out that probabty

hydroxyl ions rather than water mole-

cules participate in the diffusion process

He is proposing a mechanism involving

solid state reactions followed b y rapid

solution of the reaction products

When studying the reactions under-

gone by 3CaO.AI~O3, it must be remem-

bered that this compound can react with

water to form hydrated compounds and

with calcium sulfate, calcium hydroxide,

and other calcium salts to form double

salts Portland cements contain only a

few per cent of 3CaO.Al~O3, generally

well under 15 per cent Pastes of many

powdered cement clinkers will steam and

harden almost immediately upon mixing

of the powder and water This is gener-

ally referred to as flash set The addition

of small amounts of calcium sulfate to

these clinkers usually gives products

which, when mixed with water, produce

pastes that heat up very slowly and harden gradually This has led to the practice of calling calcium sulfate a re- tarder However, it appears from the work of Bates that the retarding prop- erties of this salt might not be great if 3CaO-SiCh did not furnish Ca(OH)~ to assist the calcium sulfate On the other hand, it appears necessary to conclude that sulfate and hydroxide function as a team in preventing flash set

Several investigators found that the early strengths of many cements were increased with increasing amounts of calcium sulfate This indicated that this salt played a role as a n accelerator as well as a retarder

Lerch (16) i n 1946 published the re- suits of a very thorough study of the influence of calcium sulfate on various properties of portland-cement mortars

I t seems very evident from that study that calcium sulfate accelerates the rate

at which either or both 3CaO.SiO2 and 2CaO.SiO~ develop strength during the early life of the cement paste It is, of course, possible that the calcium sulfate also influences the rates at which other phases contribute to strength However,

in this paper, the discussion of the accel- erating effect with respect to the develop- ment of strength will be confined to the two calcium silicates

Many studies have been made to de- termine the effects of salts, other than calcium sulfate, and other compounds upon the rates of setting and hardening

of cement pastes 3 In most of these, no effort was made to determine whether or not an observed retardation or an ob- served acceleration applied to the re- actions of 3CaO Al~Os or to those of the silicates One other confusing element in the literature on setting is that some

a I n t h i s paper, no a t t e m p t will be m a d e to

i n d i c a t e w h e t h e r n e a t pastes, mortars, or con- cretes are involved, since t h e reactions u n d e r consideration p e r t a i n to cement*water systems,

8 SYMPOSIUM ON ADMIXTURES IN CONCRETE

investigators worked with cements that

contained sufficient dehydrated gypsum

to cause what is classed as false set (1)

For these reasons, one should examine

carefully any report of work on acceler-

ators and retarders to be certain that the

author's definitions of accelerators and

retarders are clearly understood

I t seems, from this review of the liter-

ature, 4 that the investigators who at-

tempted to explain the reactions of the

cement minerals with water very quickly

reached a point where they had to en-

large their explanation to include the

formation of precipitates on the surfaces

of the crystals Phillips, Baikov, and

Roller particularly appear to have recog-

nized that the solution theory was inade-

quate Much more is known today about

surfaces and reactions at surfaces so that

it may be possible now to overcome the

difficulties of these early investigators in

formulating a satisfactory theory There-

fore, this paper is an attempt to explain

by means of solid state reactions the

mechanism by which calcium sulfate and

calcium hydroxide control flash set and

b y which calcium sulfate accelerates the

strength-producing reactions of the cal-

cium silicates The behavior of other

inorganic and organic compounds are

studied in the light of this mechanism

For simplicity, the paper will refer to

retardation of flash set and retardation

and acceleration of the rates of silicate

reactions

SOLID STATE REACTIONS

The production of portland-cement

clinker probably is entirely by solid state

reactions For example:

1 CaCO3 is decomposed into CaO and

CO2 and clay is dehydrated and decom-

4 For a comprehensive review, see H H

Steinour, "The Setting of Portland Cement A

Review of Theory, Performance, and Control,"

tories, Portland Cement Asan

posed into SiOs and AlsO3 in the solid states,

2 CaO combines with SiO2, Al~O3, and Fe~O~ in the solid state to form the solid cement minerals, and

3 After some liquid is formed by melt- ing, CaO from the liquid probably com- bines with solid 2CaO.SiO~ to form solid 3CaO Si02

Most of the industries that use cata- lysts are based on solid state reactions For example, Weyl (17) states: "We are now standing at the threshold of a new phase in chemistry, the chemistry of the solid state Recent progress in the atomic structure of solids justifies the hope that, soon, we shall be able to understand the participation of solids in chemical re- actions, including heterogeneous catal- ysis."

Some studies of solid state reactions that seem particularly important to the student of portland cement are set forth below:

Jeffery (18) discusses a number of solid state reactions and makes the following statements regarding the reaction of 3CaO SiO~ with water:

"But even in this initial period when an excess of water is present, it is not certain that the hydrate is formed by precipitation Hansen has produced evidence indicating that even in this case the reaction with 3CaO SiO~ is really a direct transformation

of the anhydrous solid and not a precipita- tion from solution During the hardening process, extending up to a year or more, it

is difficult to see how it can be anything

else."

McConnell (19), from a study of the

(B2CaO SiO2) to a gelatinous hydrated phase similar to tobermorite (3CaO 2SiO2.3H~O), found that much of the gel occurs as a homogeneous flint-like sheath, up to 5 cm thick, around nodules

I-IANSEN ON A C T I O N S OF C A L C I U M S U L F A T E 9

of unaltered larnite and he concluded,

from a detailed study of the gel, that the

progress of hydration is a solid state re-

action which is effected b y diffusion

T A B L E L - - A N A L Y S I S O F C E M E N T

P A S T E A B S T R A C T S A F T E R M I X I N G O F

C E M E N T A N D W A T E R , g p e r l i t e r

S i 0 2

R 2 0 ,

C a O

SO~

K~O

N a 2 0

7 mln after mixing Witho~ Gypsu~ 0 0 0 0.50: 1 7 3 0 0 2 0 0 7 0 0 2 2 hr after mixing With Without With GYPsum Gypsum Gypsum 0 0 0 2 0 0 0 0 0 0 0 2 0 0 0 3 0 0 0 4 0 0 0 4 2 2 2 1 4 7 2 3 0 0 9 6 0 0 3 1 0 5 0 0 0 0 5 9 0 0 6 0 0 2 0 0 0 0 0 2 Hansen (3) endeavored to show that the following reaction was a solid state reaction: 3 C a O AI~O, + CaC12 + 1 0 H g ) = 3CaO- Alg)3 CaCI2.10Hd3 (7)

A cement clinker was reburned to re- duce the SO8 content to a low value The clinker, after reburning, had the following composition in per cent b y weight: C a O 6 5 2 M g O 2 6 A1203 6 5 SO, 0 0 6 Fe~Os 3 0 K s O 0 0 3 S i 0 2 2 5 N a 2 0 0 0 9 M n O 0 5 1 This was ground to a fine powder and T A B L E I I - - - C O M P O S I T I O N O F F I L T R A T E S F R O M P A S T E S O F 200 g C E M E N T P L U S 150 m l O F CaCI2 S O L U T I O N CaCh moles per liter O 0075

0 0 1 5 0

0 0 3 0 0

0 0 6 0 0

0 1 2 0 0

Composition of Filtrate, moles per liter

N o n e ] 0 0 2 5 4 0 0 0 1 7 0 0 0 1 3 0 0 0 1 1

N o n e ! 0 0 2 5 7 0 0 0 3 9 0 0 2 7 2 ] 0 0 0 1 3 0 0 0 1 1

N o n e I 0 0 2 6 7 0 0 1 4 3 0 0 3 8 7 I 0 0 0 1 3 0 0 0 1 1

N o n e ] 0 0 2 5 5 0 0 4 4 2 0 0 6 7 2 0 0 0 1 3 0 0 0 1 1

N o n e I 0 0 2 5 0 0 1 0 1 4 ] 0 1 2 4 9 [ 0 0 0 1 3 0 0 0 1 1

Ch Combined, CaO Released, moles per liter molesper liter

0 0 0 5 8 0 0 2 1 5

0 0 1 1 2 0 0 2 3 4

0 0 1 5 7 0.O240

0 0 1 5 8 0 0 2 3 0

0 0 1 8 6 0 0 2 3 5

A number of investigators have made

quantitative analyses of extracts from

cement pastes at early periods after the

cement and water are mixed which show

the presence of only very minor amounts

of SiO2 and A1203 in these extracts For

example, Kalousek, Jumper, and Trego-

from the extract at 7 min and 2 hr from

pastes of a clinker ground with and

without gypsum

Similar data are obtained when the

extraction periods are reduced to less

than 7 min These extraction studies

give no indication that the concentra-

tions of SiO, and A1203 build up to rela-

tively high values as is common for

supersaturated solutions and then de-

crease as precipitation occurs

pastes were made of 200 g of clinker plus

150 ml of a CaCI2 solution These were shaken at room temperature for 1 rain and filtered immediately on a suction filter, which gave a time of contact be- tween the solid and liquid of a few sec- onds more than 1 min The filtrates were titrated for (OH)2 content and analyzed for SO3, C12, CaO, K20 and Na20 with the results given in Table II

The values in the next to the last column are the differences between the

Cl, content of the mixing water and that

of the filtrate This, it is believed, is the amount of Cl, that has combined as illus- trated in Reaction 7

The values in the last column are the amounts of CaO released by the cement, probably largely from the 3CaO-SiO,

10 SYMPOSIUM ON ADMIXTURES IN CONCRETE

These are obtained by subtracting the

values for CI, in the filtrate because the

CaO required by this Cls was added in

the calcium chloride

In accordance with Reaction 7, one

mole (270 g) of 3CaO.A1203 combines

with one mole of CaCls In the solution

initially containing 0.06 moles of CaCls,

0.0158 moles bad combined This indi-

cates that 270 X 0.0158 or 4.3 g 3CaO-

A120, per liter reacted with CaCI~ in a

period of slightly more than 1 rain

On the assumption that all of the CaO

released to the solution was released as

follows:

2(3CAO SiO2) + 3H20

two moles (456 g) of 3CaO SiO~ release

three moles of CaO, or two thirds of a

mole (152 g) release one mole of CaO

This indicates that 152 X 0.0230 = 3.5

g of 3CaO-SiOs per liter hydrolyzed in

slightly more than 1 min This is equiva-

lent to 0.9 g SiOs per liter that would

have been dissolved and precipitated

from the solution in this period of time

if the solution theory was correct

The 4.3 g of 3CaO-Al~O3 that com-

bined with the CaCI~ is equivalent to

1.6 g A1203 per liter It seems impossible

for these quantities of SiO, and A1203 to

go through the solution phase in a period

of 1 min when only traces of these oxides

can be found in the liquid phase It

seems necessary, therefore, to conclude

that the major portions of the calcium

alttminates and silicates in portland ce-

ment do not react with water and calcium sulfate by going into solution

Further evidence of a solid state re- action was obtained b y Hansen (20) in a study of aerated and nonaerated cement

In this study, a clinker was ground to a powder without added calcium sulfate This powdered clinker as ground and after exposure in a thin layer to labora- tory air for 24 and 48 hr was made into slurries with a solution of calcium chlo- ride, 200 g powder to 150 ml of solution containing the equivalent of 3.0 per cent Cls by weight The slurries were filtered

at 1, 5, 15, and 30 min intervals and analyzed for Cls with the results shown

in Table III

The explanation for the decreased rate

of reaction with CaCls after aeration appears to be that the surfaces of the grains of 3CaO AlsO3 reacted with HsO and COs from the air during aeration

If the reaction with calcium chloride was

a solid state reaction in which this salt

is adsorbed from the solution onto the surface of the crystals as will be explained later, it is readily understood how the previous adsorption of HsO and CO2 could interfere with this reaction I t is not easy to believe that this adsorption

of H~O and COs would so markedly re- duce the rate at which the 3CaO.A1203 crystals would dissolve

Brown (21) examined thin sections of hardened cement pastes microscopically and found that grains of CaO and of MgO had hydrated to the hydroxides without evidence of having gone through

a solution phase This hydration in situ

is used to explain the ability of these oxides to produce expansion in concrete Roller (22) and others have found the flash-setting properties of cements could

be destroyed by treating the cement with either steam or moist air

Razonk and Mikhail (23) suggest a mechanism for the hydration of MgO by water vapor which probably is similar to

HANSEN ON ACTIONS OF CALCIIYM SULFATE 11

reactions of cement minerals in cement

pastes They state: "Experiments on the

rate of hydration of magnesium oxide at

various temperatures and on the effect of

temperature on the sorption isotherms of

water vapor suggest that the initial stage

in the uptake of water vapor by the oxide

is essentially a rapid Van der Waals

physical adsorption, together with slower

chemisorpfion, and this is followed by a

diffusion process to the inside of the

solid The present investigation indicates

PROPERTIES OF SURFACES For most physical properties of a solid,

it is sufficient to assume that the prop- erties of every element of a volume within a given particle has the same properties as any other However, this assumption of homogeneity breaks down near the surface of the particle

According to Weyl (17), V M Gold- schmidt was the first to a t t e m p t a cor- relation between crystal structures and physicochemical properties of matter

r e s e n t sodium ions, large spheres represent chloride lofts

that the last stage in the hydration proc-

ess is the recrystallization of the mag-

nesium oxide water complex, formed b y

the interaction of water molecules with

magnesium oxide crystallites (probably

through O - ions) to yield the stable

magnesium oxide lattice."

In suggesting that the reactions in a

cement paste axe primarily solid state

reactions, it is, of course, realized that

some or all of the minerals will dissolve

to some limited extent This has been

demonstrated b y the analyses of extracts

from cement pastes which always con-

tain very minor amounts of SiP2 and

Al20a

He explained crystal structure in terms

of the three fundamental properties of ions: namely, their' charges, their size, and their polarizatiqn properties The m o d e m concept of crystals is that they are made up of ions For example ,

in a crystal of NaC1, a sodium ion is bonded to six chlorine ions by equal forces Attractions between ions of un- like charges and repulsions between ions

of like charges are important for an understanding of crystal structures The present concept of the atom is that it consists of a positive nucleus and

a number of planetary electrons The atomic number assigned to an atom is

12 SYMPOSIUM ON ADMIXTURES IN CONCRETE

the number of its planetary electrons

which equals the number of positive

charges in the nucleus The atomic

weight is approximately the weight of

the nucleus When an atom is converted

to an ion, it loses or gains one or more

electrons The ions in the interior of a

crystal are mutually attracted and re-

pelled by each other However, those on

the surface of a crystal are unbalanced

toward the atmosphere Weyl illustrates

this in Fig l(a) He pictures this surface

as consisting of rigid spheres with excess

electrical charges He states:

"However, the electron distribution prob-

ability of the surface ions cannot remain

spherical because the ions of high polariz-

ability, the C1- ions, must develop a higher

than average electron density towards the

interior of the crystal because their electron

douds axe attracted by the Na + ions As a

result, they exhibit lower electron density

toward space In a similar fashion, the elec-

tron distribution density of the cations

undergoes a change Each surface Na + ion

is exposed to the attraction of a C1- ion of

the lower layer but has no corresponding

anion on the opposite side which balances

this attraction As a result of this one-sided

polarization, the electron clouds of the Na +

ions are repelled towards the vacuum, a de-

formation which decreases their positive

potentials."

Weyl represents this polarized surface

in Fig l(b) The heavier lines indicate

the spots of higher electronic density of

the outer electron shell The length of

the arrow indicates the residual potential

field emanating from a surface ion into

space

The atomic number of sodium is 11

and that of chlorine is 17 The larger

C1- ion is polarized to a greater extent

than is the smaller N a + ion and, accord-

i n # y , its surface forces are weakened to

a greater extent than those of the N a +

ion b y polarization

Figure l(c) represents a rearranged

surface in which the N a + ions tend to

recede slightly so that they are more

completely surrounded b y C1- ions Weyl refers to this as a screening of the

N a + ions by C1- ions This screening gives a surface of lower surface energy than that in Fig 1 (b) This surface has also become predominately negative whereas that in Fig 1 (b) was predomi- nately positive

Polarization and screening as illus- trated in Fig 1 for NaC1 must be im- portant in the reactions of cement in the water In dealing with a material made

up of powdered crystals such as portland cement, the freshly produced surface might tend to exhibit the characteristics

of Fig l(a) but probably would very quickly assume the characteristics of either (b) or (c)

Much has been written (z4) about de- fects in crystals which play a prominent part in solid state reactions These m a y play an important role in the reactions

in cement pastes and possibly the studies that are being made in a number of laboratories on the structures of the cement minerals will eventually throw light on this role Also, the particles of cement, except for the very finest, are agglomerates of m a n y fine crystals The following quotation from Weyl appears pertinent in this connection: "Can a polycrystalline aggregate of an oxide ever be considered to be a precise stoichi- ometric ratio, or are all grain boundaries and internal surfaces the seat of 'avail- able electrons' and do they represent a potential source of O - ions? Recent ex- perimental work led us to suspect that even apparently stable substances can undergo an exchange of electrons with adsorbed 02 molecules which leads to the formation of atomic oxygen Grain boundaries, interfaces, and surfaces are the seats of these reactions."

SURFACES OF CALCIUM ALUMINATE AND SILICATE CRYSTALS

In accordance with the above discus- sion, the surface of a calcium aluminate

HANSEN ON ACTIONS oF CALCIUM SULFATE 13

crystal would consist of Ca ++, AI: : : and

O- ions and that of a calcium silicate

crystal would consist of Ca ++, Si I l j l

and O- ions However, the smaller and

less numerous A1 (IE and Si 'l l l l ions,

atomic numbers 13 and 14 respectively,

would likely be screened by the larger

and more numerous Ca ++ ions and by

the more numerous O- ions Hence, the

ions in the surfaces of 3CaO.AI~O,,

3 C a 0 S i O , , and 2CaO-SiO, are prob-

ably principally Ca ++ and O- ions and

will be so considered in the discussions

which follow

If these surfaces were either polarized

or rearranged to the extent that they

were either predominately negative or

positive, the particles would repel each

other Steinour (2S) and others have

shown that particles of cement in water

tend to attract each other and form flocs

Hence, it appears that the surfaces Of the

powdered crystals of the cement minerals

carry approximately equal amounts of

positive and negative charges

When either of these surfaces is ex-

posed to water, the positive ions Ca ++

should attract the O H - ions of the water

and the O- ions should attract the H30 +

ions of the water As Razouk and Mikhail

suggest for MgO, the first step may be a

physical adsorption followed by chemi-

sorption and then by a diffusion of O H -

and H30 + ions into the crystal

Physical adsorption, as suggested by

Razouk and Mikhall, must be a dynamic

process in which a Ca ++ ion may tem-

porarily unite with a negative ion and

then release it to the solution and unite

with another negative ion If these ad-

sorbed ions are able to migrate into the

crystal and cause a rearrangement of the

crystal, the process becomes chemisorp-

tion which means that the ions have

reacted chemically with the solid surface

This chemical reaction, as in the case

cited by Razouk and Mikhail, has taken

place without the solid going into solu-

as that of G A Mills and S G Handin, with 018, shows clearly that there is a con- tinuous exchange of oxygen ions of the sur- face layer of an oxide and the oxygen molecules of the atmosphere."

The rate at which the diffusion into the crystal takes place should depend on the crystal structure of the compound Weyl suggests that crystal structure can ex- plain the difference between the reactiv- ity toward water of ~ and 72CaO.SiO, and Jeffery suggests that the difference between the reactivities toward water of B2CaO-SiO2 and 3CaO SiO, may be ex- plained b y differences in crystal struc- ture Foreign atoms, such as MgO and A1,O, in the 3CaO.SiO: crystal, may markedly affect the reactivity of the crystal in a solid state reaction

One of the common arrangements of SiO~ in crystals is the SiO4-tetrahedron and the formula for an orthosilicate such

as 2CaO.SiO2 is usually written as CasSiO4

Jeffery (3) says regarding the pseudo- structure of 3CaO-SiO~ : "The structure

is built from discrete SiO4-tetrahedra together with separate oxygen and cal- cium ions the main features of the structure of 3CaO SiO~ are the irregular coordination of the calcium ions and the 'holes' adjacent to these ions Since ir- regular coordination gives rise to dis- tortion of the electrostatic field between ions, 3CaO SiO~ necessarily has a high lattice energy."

Midgley (20) makes the following state-

14 SYMPOSIUM ON ADMIXTURES IN CONCRETE

merits regarding the structure of B2CaO-

SiO2 :

"The structure is built up of isolated

SiO4-tetrahedra and calcium ions Although

the coordination of calcium ions is irregular,

there are no open spaces as in 3CaO-SiO~

which would allow ready attack by hydroxyl

ions This would account for the fact that,

although both compounds have valuable

cementing properties, 3CaO-Si02 is very

rapidly hydrated and 2CaO.SiO~ only

slowly."

The m a n y investigations (18) on the

CaO-SiO,-H,O system have not as yet

clearly defined the compositions of the

hydrated calcium silicates that exist in

the normal hardened pastes of mortars

and concretes Some (27) believe that the

equilibrium compound is what m a y be

classed as tobermorite, 3CaO-2SiO,

3H~O There is some evidence (27) that

afwiUite, 3CaO 2SiO2.3H,O, might be

formed in these pastes and there is con-

siderable evidence (18) that the product

might be 2CaO SiO, H20

Heller (~-8) studied the crystal structure

of 2CaO SiO, a hydrate which is formed

at elevated temperatures This work in-

dicated the silicon formed ( S i O ~ ( O H ) - - -

tetrahedra in this compound and Megaw

(29) concluded that the silicon occurred

in the same manner in afwillite Heller

concluded that the formula for the

2CaO SiO~ a hydrate was Ca2(SiO3OH)

OH and Megaw concluded that the

formula for afwillite was Caa(SiOaOH)2"

T h a t is, a Ca ++ ion in the surface of

the crystal could chemisorb an O H - ion

which could then diffuse into the crystal

An O - ion in the surface could chemi-

sorb an HsO + ion and form SiO3OH

H , O - - - from which the H30 + ion could diffuse into the crystal Both of these chemisorption processes could be re- peated over and over without the 2CaO SiO~ going into solution but with a grad- ual rearrangement of its crystal structure The reactions to form Ca3(SiO3OH)2" 2H20 could be as follows:

Ca ++ + 2OH- = Ca(OH), 2(SiO,O) + 2H,O +

= 2(SiO, O H H s O ) - - -

In this reaction, one Ca ++ ion in the surface of the crystal chemisorbs two

O H - ions and splits off as C a ( O H ) , Two

O - ions chemisorb 2H30 + ions to form

sorbed H30 + ions couM diffuse into the crystal and change the structure of the crystal

These reactions need not be limited to chemisorption of ions However, it seems that, in solutions where ions are present, ions would be chemisorbed in preference

to molecules of water However, when crystals of such unstable compounds as 3CaO-Al~O3 and 3CaO-SiO~ are exposed

to water vapor, they must react rather rapidly with water as is indicated b y con- siderable experimental evidence Typical reactions might be as follows:

C a ++ + H s O ~ Ca(OH)2 SiOv O - - - - + H~O = SiOv O H - - - + OH-

Ca ++ + OH- = CaOH + One reaction suggested for 3CaO SiO~

in cement pastes is:

2(3CaO.SiO2) + 6HzO

= 3CaO-2SiO2.3H,O + 3Ca(OH)~ (8)

If this reaction occurred only by the anhydrous silicate dissolving congruently and reprecipitating as the two reaction products, one would expect the hydrated silicate to have a uniform composition Also, one would expect, if the paste

HANSEN ON ACTIONS OF CALCIUM SULFATE 15

reached a stage in which liquid water no

longer existed, that the reaction would

stop and leave unreacted crystals of

3CaO.SiO, These crystals could con-

tinue their reaction later whenever liquid

water could contact them

In the suggested solid state reaction, it

is visualized that the reaction could con-

tinue in the absence of free water, which

seems to be clearly established, and that

the composition of the hydrated calcium

silicate reaction product might vary

widely from what was originally the out-

side of a crystal of 3CaO-SiO2 to the

center of that crystal The degree to

which this product assumes a well de-

fined crystalline structure would be ex-

pected to vary It can also be visualized

that Ca(OH)2 that splits out of the sili-

cate could exist physically within the

hydrated silicate in a finely divided

amorphous state Similar reactions would

be expected for the other compounds,

such as 3CaO-Al2Os and 2CaO.SiO2

In attempting to give a picture of the

mechanism of setting and hardening of

a cement paste, it is necessary to use

equations to illustrate reactions Reac-

tions such as 5 and 8 are written to in-

dicate what might happen to crystals of

3CaO.AI203 and 3CaO SiO, in a normal

concrete In such concrete, sufficient

OH-, HaO + and SO4 ions to destroy

the crystal structures of the anhydrous

crystals might migrate into the crystals

in relatively short periods of time but it

might require much longer periods of

time for a sufficient number of molecules

of water to diffuse into those partially

reacted crystals and cause the complete

rearrangement of the structures to those

of compounds such as ettringite and

tobermorite as indicated by Reactions 5

and 8 T h a t is, one may visualize a hard-

ened cement paste to be a system which

is tending toward a certain equilibrium

but which may never reach that equilib-

rium in normal concrete

It is not possible at this time to present these reactions as more than possible explanations of the solid state reactions that might be responsible for the setting and hardening of cement pastes How- ever, it does seem, from what is known about reactions at surfaces and from the evidence that indicates that the cement minerals do not dissolve at rates required

by the rates at which cement pastes set, that it is necessary to conclude that the setting and hardening are caused by the chemisorption and migration of O H - and H30 + ions from water into the crystals

of the cement minerals If this is true, then the problem of controlling the rates

of these reactions becomes one of con- trolling the concentrations of O H - ar/d H30 + ions available to the surfaces of the cement particles'and by controlling the amount of surface available to those ions The latter has, of course, been found

to be effective by changing the fineness

1 A12Os This indicates that the first reaction of 3CaO.A1203 with water is

3CaO-AI20~ -4- 9H~)

= 2CaO'AI~DVSH~O + Ca(OH)l (9)

In a cement paste with a limited amount of water, this reaction probably takes place as follows:

Ca ++ + 2(OH)- = Ca(OH)~ (10) O- + H 3 0 + = H 2 0 + O H - (11)

T h a t is, a Ca ++ ion in the surface of the crystal chemisorbs two O H - ions and

16 SYm, osrtr~ ON ADmX~YRES m Coscm~Tv

splits off as Ca(OH)2 and an O- ion

chemisorbs an HsO + ion to form H g )

and an O H - ion The H20 + O H - then

diffuse into the crystal and the chemi-

sorption processes are repeated

As the concentration of C a ( O H h

builds up in the liquid phase of the paste,

the 2CaO-AbO~-8H20 formed on the

surface of the crystal of 3CaO AI~O3 can

chemisorb Ca ++ and O H - ions to form

either 3CaO Al~O3.6H~O, 3CaO AI~a-

3Ca(OH)~.3OHzO The extent to which

these compounds can be formed is de-

pendent on Ca(OH)~ released in Reaction

9 unless Ca(OH)~ is furnished from other

sources, such as the hydrolysis of 3CaO

SiO~

Phillips observed that, if a paste of

3CaO.Al~O3 and water that had heated

up and stiffened, was reworked, it would

heat up and stiffen at a slower rate This

indicates that the Ca(OH)~ released in

accordance with Reaction 9 during the

first period is capable of retarding the

rate of further reaction This also indi-

cates that the formation of 3CaO A l ~ 8

Ca(OH)2.12H20 on the surfaces of the

crystals of 3CaO.Al~Oa, as postulated

b y Roller, cannot be the mechanism by

which flash setting is prevented

The difference between the original

and the reworked paste is that the liquid

phase of the latter contained dissolved

Ca(OH)2 which the former did not This

Ca(OH)~ represses the ionization of water

and the concentration of the H30 + ions

is much lower in the liquid phase of the

reworked paste than in that of the orig-

inal paste

If the reaction of Reaction 9 is a solid

state reaction, its rate depends upon the

rate at which O- ions can chemisorb

H30 + ions Since the liberation of

Ca(OH)9 by 3CaO.A12Oa can retard the

rate of the reaction, it seems evident

that the concentration of H30 + ions is

the governing factor and that the re-

action of Reaction 9 is the reaction responsible for flash set One might then ask whether or not the addition of CaSO~

to the mixing water can retard the rate

of this reaction

The addition of calcium sulfate to water has very little effect upon the ionization of water Hence, from this standpoint, it should neither retard nor accelerate the rate of this reaction How- ever, the S O c - ion is fairly large and the adsorption of it by the Ca ++ ions in the surface of the crystal would interfere with the adsorption of O H - ions The adsorption of Ca ++ ions from the solution

by O- ions could also interfere with the adsorption of H / 3 + ions There appears

to be no way at present of determining the influence of these factors upon the rate of this reaction, but one might ex- pect that this would not be very great unless the S O c - ions were very effec- tive in screening the O - - ions Since Bates found very little retardation by CaSO4, it appears that the S O c - ion is not very effective in preventing chemi- sorption of H30 + ions by the O - ions However, when solid calcium sulfate is mixed with solid 3CaO.Alg)8, the salt cannot be effective until it gets into solu- tion and, by the time an appreciable amount has dissolved, a considerable re- action between the 3CaO.Alg)8 and H:O may have taken place Hence, the results obtained b y Bates with pastes with only minor admixtures of either CaSO4 or Ca(OH)~ or blends of the two may not give a true picture of the retard- ing powers of these materials

When a cement is flash setting, the paste heats up and stiffens very markedly within one or two minutes after the ce- ment and water are mixed It is generally believed that gypsum will prevent flash set in most cements This belief is based

on results with cements in which gypsum was ground with the clinker but it neg- lects the fact that, in many of the ce-

HANSEN ON ACTIONS OF CALCIU~ SULFATE 17

ments, some or all of the gypsum prob-

ably was converted to either hemihydrate

or soluble anhydrite either during grind-

ing or storage There are some clinkers

that will produce flash setting cements

when ground with gypsum but will pro-

duce normal setting cements when

ground with a blend of gypsum and

hemihydrate

As pointed out earlier, most of the

alkalies enter the mixing water as sulfates

and not as hydroxides as believed by

Roller Most clinkers combine with SO3

at a fairly rapid rate As stated pre-

viously, Hansen showed that one clinker

combined with 0.22 g CI~ per 100 g of

cement in 1 min, which was calculated

to be equivalent to 0.74 per cent SOs by

weight of cement Roller made his ex-

tractions 15 min after the paste was made

and his analyses showed very little SOs

in those extracts He, therefore, assumed

that the alkalies had dissolved as hy-

droxides whereas they very likely dis-

solved as sulfates because his clinkers

contained from 0.23 to 1.02 per cent SOs

Most of the hydroxide required for

Roller's Reaction 1 comes as Ca(OH)2

from the hydrolysis of the cement min-

erals, such as 3CaO.AI~)3 and 3CaO

SiO2 I t is true that the calcium sulfate

would tend to permit a larger amount of

Ca(OH), to remain in solution than

would be possible if the solution was pri-

marily a solution of alkali hydroxides

instead of alkali sulfates However, if the

Ca(OH)~ is dissolved from the 3CaO

SiO, for example, it would be available

to react with 3CaO A1203 before it pre-

cipitated out as solid Ca(OH)~ The

point that Roller and others appear to

overlook is that the reaction or reactions

that cause flash set are proceeding rap-

idly within seconds after the water comes

in contact with the cement

I t does not appear possible for 3CaO

AlcOa to furnish the Ca(OH)2 required

to form a coating on itself of 3CaO-

Alg)3.Ca(OH)~-12H20 Hence, it ap- pears, if such a coating is required to prevent flash set, that the Ca(OH)2 has

to come from the hydrolysis of 3CaO SiO2 In that case, one would have to explain why CaSO4 accelerated the rate

of hydrolysis of 3CaO-SiO2 This sub- ject will be discussed later However, it does not seem likely that the retarding power of CaSO4 lies solely in its ability

to accelerate the rate of hydrolysis of 3CaO SiO2 or of any of the other cement minerals

I t seems necessary, therefore, to return

to the conclusion held by many inves- tigators to the effect that some combina- tion of SOa with 3CaO.A12Oa prevents flash set I t is well known that SOa can combine very rapidly with 3CaO.AI,O3 without causing flash set This indicates that Roller was wrong when he postu- lated that the formation of 3CaO.Alg)3 3CaSO4.32H~) caused flash set How- ever, as pointed out above, there are cases in which a cement that combines rapidly with SO3 shows flash set when the liquid phase of the paste is not sat- urated with respect to gypsum but which shows no flash set when the liquid phase

is supersaturated with respect to gypsum This indicates that some other reaction

of 3CaO- Al~03 than the formation of one

or the other sulfoaluminate is responsible for flash set and supports the conclusion that Reaction 9 is that reaction

If this is true, then what property of CaSO4 is involved? Since increasing the quantity of CaSO4 in solution will pre- vent flash set, it seems that the dissolved CaSO, must interfere with this reaction

If the rate of the reaction producing flash set depends upon the rates at which the

Ca ++ and O- ions in the surfaces of the crystals chemisorb O H - and H30 + ions, then calcium sulfate can interfere with this reaction by the Ca ++ and S O 4 - - ions of the calcium sulfate being physi- cally adsorbed on the surface Ca++ and

18 SYMPOSIUM ON ADMIXTURES IN CONCRETE

O- and blocking these sites from O H -

and HsO + ions Calcium sulfate hemi-

hydrate and soluble anhydrite are two

compounds that, when present to the

extent of a few per cent in a cement

paste of normal water content, can pro-

duce relatively high concentrations of

CaSO4 in the liquid phase in a few sec-

onds because they are unstable in water

So far as one can determine, the first

use of calcium sulfate for preventing

flash set was the addition of plaster of

Paris to the mixing water This was fol-

lowed b y grinding the clinker with plaster

of Paris and later by grinding with gyp-

sum However, in the latter, the gypsum

probably was converted to plaster of

Paris either during grinding or storage

If the first investigators had added

powdered gypsum to the mixing water,

they very likely would not have been

satisfied with the performance of their

concrete because their cements probably

contained relatively large amounts of

3CaO AltOs and the addition of gypsum

probably would not have prevented flash

set

Flash setting can be eliminated from a

clinker by permitting the clinker to ab-

sorb moisture Hanscn showed that this

treatment greatly reduces the rate at

which the clinker reacts with calcium

chloride and sulfate This indicates that

reactive sites on the surfaces of the

clinker have been blocked by chcmisorp-

tion of moisture from the air It seems

that it would be impossible to form

chcmisorption of water by 3CaO-Al~O~

and, of course, no sulfoaluminate can be

formed on the surface Hence, something

besides the formation of these double

erties of the clinker

From the standpoint of flash set, port-

land-cement clinkers vary widely Some

show no signs of flash set when tested

without an admixture, some have mild

flash setting tendencies that are easily

controlled by small additions of gypsum, and others have strong flash setting ten- dencies that can only be controlled by the addition of either hemihydrate or soluble anhydrite It is conceivable that salts other than CaSO4 might be effective in preventing flash set with some clinkers and not with others

From the standpoint of the solid state reactions proposed above, one might predict that any salt with one large ion might be effective to some degree, pro- vided it did not react in some other way with the cement minerals Soluble mag- nesium salts, for example, would react with Ca(OH)a and precipitate insoluble Mg(OH)2 and yield a calcium salt Alkali phosphates would react with Ca(OH)~ to produce insoluble calcium phosphates and alkali hydroxides Salts of these types, Forsen classified as destroyers All evidence points to the conclusion that the reaction of 3CaO-AI203 with water starts at an extremely high rate the moment that water comes in contact with the crystals Hence, any material that is going to exert a retarding effect upon this reaction must be able to dis- solve at an extremely rapid rate and form ions that are adsorbed by the O- ions sufficiently to block most of the O- ions against the adsorption of H30 + ions Plaster of Paris and soluble anhydrite are two compounds that can dissolve at such rates and 3CaO SiO~ is a compound from which Ca(OH)~ can hydrolyze at a rapid rate I t appears that this combination of

an unstable calcium sulfate and 3CaO SiO~ is required to prevent flash set in

amounts of 3CaO.AI~O3 From this, it seems evident that it would be unsafe to generalize regarding the retarding powers

H A N S E N ON ACTIONS OF CALCIUM SULFATE 19 ate the rates at which cements develop

strengths This, it seems, must be due to

effects that these salts have upon the

rates at which 3CaO-SiO2 and possibly

2CaO.SiOz react with water Forbrich

(30) showed that calcium chloride in-

creased the rate at which a paste of

3CaO SiO2 released heat

As discussed under properties of sur-

faces, these silicates may react to form

minerals, such as 2CaO.SiO2-2H~O;

afwillite, 3CaO.2SiOv3H43; and tober-

morite, 3CaO.2SiO2.3H20 These re-

actions are explainable as follows:

C a ++ + 2 O H - = C a ( O H ) 2 (10)

C a ++ + O H - = C a O H - ( l l )

( S i O , - O ) + H 3 0 +

= (SiO3OH.H20)- - - (12)

In Reaction 10, a Ca ~- ion has com-

bined with two O H - ions to split out a

molecule of Ca(OH)2 In Reaction 11, a

Ca ++ ion chemisorbs an O H - ion which

can then diffuse in the crystal to other

Ca ++ ions In Reaction 12, an O- ion in

the surface of a silicate chemisorbs a

H30 + ion to form an (SiO3OH.H~O) -

ion The adsorbed H30 + ion can then

diffuse to other O- ions

Presumably, Reactions 10 and 11

could not progress to an appreciable ex-

tent unless Reaction 12 occurred simul-

taneously or vice versa Hence, these

reactions are controlled by the concen-

trations of O H - and H30 + ions in the

liquid phase of the cement paste Since

3CaO.A1203 and 3CaO-SiO2 split off

Ca(OH)s, the liquid phase of the cement

paste almost immediately contains an

abundance of O H - ions These ions, al-

though required for Reactions 10 and 11,

act as retarders for Reaction 12 by de-

pressing the ionization of water There-

fore, to increase the rates of Reactions 10,

11, and I2, it is necessary to offset the

depressing effect of the O H - ions in the

ionization of water

Roller removed extracts from pastes

of clinkers with and without added cal- cium sulfate 15 min after the pastes were made The O H - equivalents expressed as milliequivalents per ml were 0.2320, 0.2638 and 0.3860 for three clinkers and 0.0730, 0.1030, and 0.1150, respectively, for the clinkers with added calcium sul- fate These data demonstrate that salts will have the effect of increasing the con- centrations of H30 + ions and, accord- ingly, should have the effect of accelerat- ing the rates at which the silicates develop strengths

Other calcium salts should have this ability shown by calcium sulfate to pre- cipitate calcium hydroxide from solu- tions of alkali hydroxides and, accord- ingly, increase the concentration of H30 + ions in the solution All such salts should act as accelerators unless their anions could block the surfaces of the crystals against the chemisorption of H30 + ions This will be discussed in more detail in the next section

One might expect that S O c - ions should screen much more effectively than the C1- ions and that CaC12 would be a better accelerator than CaSO, This seems to be the case However, since CaCI, is much more soluble than CaSO4,

it might tend to depress the solubility of Ca(OH), more than the latter does and

be a more effective accelerator because

of that

ORGANIC MODIFIERS

plication of dispersing agents in the ce- ment industry developed from research

in a program of developing dispersing agents for pigments in aqueous suspen- sions This led to the discovery that cer- tain polymers of condensed naphthalene sulfonic acid, when used in amounts of approximately 0.1 per cent by weight of cement, would almost completely dis- perse the floes of cement particles ob- served in cement pastes when viewed on

a slide under a microscope These studies

20 SYMPOSIUM ON ADMIXTURES IN CONCRETE

were followed by studying derivatives of

lignin and finding that salts of lignosul-

fonic acids were effective dispersing

agents Other experiments demonstrated

that a dispersing agent, when added to

the mills used in grinding cement, acted

as a "grinding aid" by increasing the

fineness of cement at constant rate of

grinding or by increasing the rate of

grinding at constant fineness They also

showed that the amount of water re-

quired by a given concrete of a given

slump could be reduced when a dispers-

ing agent was added to the concrete This

led to referring to these products as

"water reducers." These products re-

tarded the rate at which concrete de-

veloped strength and, accordingly, be-

came known as "retarders." From this,

it is seen that certain agents which can

counteract the natural flocculating proper-

ties of the powdered cement minerals

may be used as a grinding aid, as water-

reducing agents and as retarders The

following is an attempt to explain the

mechanisms whereby water-reducing ad-

mixtures and set-retarding admixtures

function

It should be pointed out that these ma-

terials, except when used as grinding

aids, were always used with cements,

clinker, plus calcium sulfate Hence,

classifying them as retarders did not im-

ply that they were useful in preventing

flash set Limited experience has shown

that the small amounts normally used

for dispersion and a moderate retarda-

tion of the rate of developing strength

will not prevent flash set with some

clinkers in the absence of calcium sulfate

Forbrich (30) found that a dispersing

agent added to a paste of 3CaO.Ald),

plus 12 per cent gypsum delayed the rate

of evolution of heat He found the same

effect, when the dispersing agent was

added to a paste of 3CaO SiO2 that con-

tained no calcium sulfate He found that

calcium chloride had very little effect

on the paste of 3 C a O A l g h plus calcium sulfate but increased the rate at which heat was liberated from the paste of 3CaO SiO~ Steinour (3) has shown that many organic compounds retard the rate

at which pastes of 2CaO.SiO~ develop strength

Ernsberger and France (32) studied the electrophoretic migration under a microscope of particles of cement sus- pended in water and in a solution of cal- cium lignosulfonate Cement particles suspended in distilled water showed no tendency to migrate toward either elec- trode but they tended to flocculate

In the solution of calcium lignosulfonate, the cement particles tended to be dis- persed and to migrate to the anode This indicated that they had acquired nega- tive charges by adsorbing negative ions from the solution The existence of such anions was demonstrated by applying a potential to a solution of calcium ligno- sulfonate in an electrophoresis cell which caused the brown lignosulfonate complex

to move toward the anode and a precipi- tate of calcium hydroxide to form at the cathode

Ernsberger and France determined the adsorptions by portland cement of the calcium salts of three lignosulfonic acids

of different molecular weights At con- centrations of 0.02 g per liter, the ad- sorptions in mg per g of cement were approximately 13.0, 4.2, and 1.8, respec- tively, for the acids having average mo- lecular weights of 9500, 2140, and 250 to

920

These demonstrations led to the con- clusion that the adsorption of the large anion was responsible for dispersion, water reduction, and retardation How- ever, no effort was made to explain the mechanism by which the anion operated The oil well cement industry has used organic materials, not for the purpose

of preventing flash set but of prolonging the perio-I during which a cement glurry

nANSEN ON ACTIONS OF CALCIUM SULFATE 21

remains sufficiently fluid for transporting

by pumps at elevated temperatures

Hansen (3) pointed out that the organic

materials used for this purpose generally

contained one or more CHOH groups

and Steinour (3) suggested that the ac-

tive group was the OH group He pointed

out that compounds with OH groups at-

tached to benzene rings, as well as some

inorganic acids containing OH groups,

were effective retarders This led him to

the conclusion that these compounds

were adsorbed by means of hydrogen

bonding on the surfaces of the crystals

of the cement minerals

The principal organic materials used

as water-reducing and retarding agents

at normal temperatures are either the

acids or the salts and derivatives of ligno-

sulfonic and hydroxylated carboxylic

acids Chemists are not in complete

agreement regarding the structures of

lignin compounds but cement technolo-

gists probably are safe in visualizing

lignosulfonic acids as being composed of

several of the following units:

HOCI~0CrM,1CHsCHOCHHS0z

This unit contains an OH group at-

tached to a benzene ring and an HSOs

group The latter is a relatively strong

acid and ionizes to yield H80 + ions and,

accordingly, can combine with cations

to give the MSOs group, where M repre-

sents a cation For simplicity, a ligno-

sulfonic acid salt might then be written

(HORSOs-), + XM + ions For sim-

plicity, we may consider X to be 2

The negative ions of these salts could

react as follows by ionic bonding with

the Ca ++ ions in the surfaces of the ce-

ment minerals:

Ca ++ + (OaSROH-)s = Ca(O,SROH)~

In hydrogen bonding, an 0 reacts as

follows with an OH group:

O - + (HORSOvM)2 = O(HORSO~M)~ There are a number of hydroxylated carboxylic acids Tetrahydroxy adipic acid, for example, has the following for- mula:

(HO)4C4H4(COOH),

The negative ions of salts of this acid could react as follows by ionic bonding with Ca ++ ions in the surfaces of the cement minerals:

Ca ++ + (OOC)vr~C,(OH),

= Ca(O0C)~I~C4(OH) They could react as follows by hydro- gen bonding with O- ions:

O- + (HOhC~H,(COO h

= O(HOhC~h(COOh The unionized molecules couM react

as follows by hydrogen bonding with O- ions:

O- + (HO)4C.I~(COOM)2

= O(HO).C~I4(COOM) 2 These hydroxylated carboxylated acids may undergo what is known as chelation

whereby an OH group in one molecule may react as follows with an OH group

in another molecule:

ROH + OHR ROHOHR

A number of molecules may be combined

in this way to give very large molecules Weyl has shown how the ions in the surfaces of crystals may be rearranged

to lower the surface energy I t is not

22 SVM_POS~VM ON ADMIXTtrREs IN CONCRETE

possible with our present knowledge to

predict the arrangements of adsorbed

ions and molecules on the crystals of the

cement minerals However, it is known

that some of these are arranged to give

particles that repel each other Also, it

is known that certain adsorbed ions and

molecules retard the rate at which ce-

ment pastes stiffen If the rate of the re-

action shown in Reaction 12 controls the

rate at which the cement reacts with

water, then the adsorbed ions and mole-

cules must retard the rate at which H30 +

ions can be adsorbed by O- ions in the

surfaces of the cement minerals

A large ion such as (HORSO~-), ad-

sorbed on Ca ++ ions might screen the O-

ions and in that way retard the rate at

which they can adsorb H30 + ions The

adsorption by O- ions through hydrogen

bonding of ions such as (HORSO3-), or

(I-IO)4Cd-I4(COO-)2 or of molecules such

(COOM)2 would block the O- ions from

H30 + ions These adsorbed ions and

molecules might go through some rear-

rangement by the action of polarizing

forces or they themselves might adsorb

other ions or molecules through hydrogen

bonding

It is not within the scope of this paper

to attempt to differentiate either be-

tween classes of compounds or between

members of a given class in their ability

to influence dispersion, water reduction,

and retardation in cement pastes How-

ever, it seems reasonable to believe from

what may be expected regarding the ad-

sorption of different types of ions and

molecules that the characteristics of the

surfaces produced by the adsorptions of

different classes of compounds on the

cement minerals will differ and that these

differences will be reflected in properties

of the cement pastes

The retarding effect of these organic

retarders appears to diminish with time

Flint and Wells (a4) showed that 3CaO

A1203 could form the compounds 3CaO AI~O3.3CaSiO3.30 to 32H~O and 3CaO AI~O3.CaSiO~-12H20 and that 3CaO

3CaO A1203.3Ca(OH)~ 30H20 as well as the compound 3CaO.AI~Os.Ca(OH)s- 12H20 They classified the compounds

in which one mole of 3CaO-AI~O3 com- bines with 3 moles of another salt as type

I compounds, and the others as type I i compounds, and listed formulas for ten

of the type I and for nine of the type I I compounds, taken from work of other investigators It seems, from the work reviewed by Steinour O), that most of the type I compounds will form complete series of solid solutions with each other and that most of the type II compounds will form complete series of solid solu- tions with each other

It appears from these studies on these type I and I I compounds that 3CaO- AI20~ will form double salts with all calcium salts Hence, calcium salts of both organic and inorganic acids should

be removed from the liquid phase of a cement paste by forming a double salt with 3CaO.AI203

Lerch, Ashton, and Bogue (8) showed that the sulfoaluminate formed in pref- erence to the chloroaluminate in a mixed solution of calcium sulfate and chloride

It seems that the least soluble double salt should be formed first Hence, if the compound formed with the calcium salt

of an organic acid was less soluble than the sulfoaluminate, it should form in preference to the latter If this occurred, the organic salt nlight be removed rather rapidly from the liquid phase of the ce- ment paste and, accordingly, exert its retarding power for a relatively short period of time It would seem, therefore, that the length of time during which such a retarder would function would de- pend upon the 3CaO.AI~O8 content of the cement Tuthill and Cordon (35) found that 0.5 per cent of a retarder pro-

H A N S E N O N ACTIONS OF CALC13~ SULI~ATE 23 longed the setting time of one cement

from 6 to 11 hr and of another from 6

to 14 hr Lukyanova, Segalova, and

Rehbinder (3~) found that the adsorption

of sulfite liquor by cement was a func-

tion of the 3CaO.Al~Os content of the

cement and was an irreversible process

In some cases, it is desirable to use a

dispersing agent to reduce the amount

of water required for a given cor~rete

without retarding greatly the rate of at-

taining early strength In such cases, a

mixture, for example, of an organic re-

tarding agent and calcium chloride may

be used as the admixture The large nega-

tive ions and molecules of the dispersing

agent are adsorbed and retard the rates

of the reactions However, as its concen-

tration is reduced b y reaction with the

3CaO.A1203, the added CaCI~ can func-

tion to increase the rates at which the

silicates develop strength This increase

might equal or exceed the retardation

produced during the life of the dispersing

agent

S ~ A R Y

A review of the literature leads to the

conclusion that many investigators of the

setting and hardening of portland ce-

ment pastes recognized that the cement

minerals reacted with water without dis-

solving in the liquid phase of the cement-

water paste

This paper considers the reactions in

such pastes in terms of 3CaO-A1203,

3CaO- SiO~, and 2CaO SiO2 It suggests

a mechanism for these reactions based

on the chemisorption of O H - and H30 +

ions from water b y Ca ++ and O- ions in

the surfaces of the crystals in which the

chemisorbed ions diffuse into the solid

crystals and transform them into solid

reaction products In this mechanism, 3CaO.A1203 causes flash set by the fol- lowing reaction:

3CaO.Alg)s + 9H~O

= 2CaO-AI~O,.SHg) q- Ca(OH)2

Calcium salts accelerate the rate at which pastes of the calcium silicates de- velop strength by decreasing the concen- tration of Ca(OH)2 in the liquid phase of the cement paste b y the following reac- tion which causes Ca(OH)~ to be pre- cipitated:

2MOH -4- CaR ffi M~.R -[- Ca(OH)2

where M is either or both Na and K, and

R is an anion

By decreasing the amount of Ca(OH)2

in solution, the concentration of H30 + ions is increased and thereby the rate

at which these ions can be chemisorbed

by O- ions

The dispersing, water-reducing, and retarding properties of organic materials are explained on the basis of adsorption

of the organic anions and molecules A large anion adsorbed by the Ca ++ ions

in the surfaces of the crystals will tend

to block the O- ions in those surfaces and, accordingly, retard the rates at which they can chemisorb H30 + ions The charges in the adsorbed anion may

be rearranged in the surface to produce surfaces which are predominantly nega- tively charged and, accordingly, repel each other Similarly, the O- ions may adsorb by hydrogen-bonding anions and molecules that contain OH groups These will block the O- from HsO + ions and may undergo rearrangement to produce surfaces that are predominantly nega- tively charged

24 SVMPOSIUM ON ADMIXTURES IN CONCRETE

R E F E R E N C E S

(1) W C Hansen, "The Properties of Gypsum

and the Role of Calcium Sulfate in Port-

land Cement," ASTM B u I a z ~ , No 121,

Feb 1956, p 66 (TP54)

(2) G A Rankin and F E Wright, "The Ter-

nary System CaO-Al2Oa-SiO2 ," The Ammi-

can Journal of Science, Vol 39, No 1

(1915)

(3) The Third International Symposium on the

Chemistry of Cement, Proceedings, London

0952)

(4) P H Bates and A A Klein, "Properties

of Calcium Silicates and Calcium Alumi-

nares Occurring in Normal Portland Ce-

ment," Nat Bureau of Standards, Technical

Paper No 78 (1917)

(5) P H Bates, "Cementing Qualities of the

Calcium Aluminates," Nat Bureau of

Standards, Technical Paper No 197 (1921)

(6) A J Phillips, "Colloidal Tricalcium

Aluminate," Journal, Am Ceramic SOc.,

Vol 2, p 708 (1919)

(7) P S Roller, "The Setting of Portland

Cement Chemical Reactions and the

Role of Calcium Sulfate," Industrial and

Enginezring Chemistry, Am Chemical Soc.,

Vol 26, p 669 (1934)

(8) W Lerch, F W Ashton, and R H Bogne,

"The Sulfoaluminates of Calcium," Journal

of Research, Nat Bureau of Standards,

Vol 2, p 715 (1929)

(9) H H Steinour, "Aqueous Cementitious

Systems Containing Lime and Alumina,"

Bulletin 34, Portland Cement Assn.; Wil-

helm Eitel, "Recent Investigations of the

System Lime-Alumina-Calcium Sulfate-

Water and Its Importance in Building Re-

search Problems," Journal, Am Concrete

Inst., Vol 28, p 679 (1957)

(10) G L Kalousek, C H Jumper, and J J

"Pregoning, "Potassium Sulfate in Cement

CFmker," Rock Products, April, 1941, p 52

(11) L Forsen, "The Chemistry of Retarders

and Accelerators," Symposium on the

Chemistry of Cement, Stockholm (1938)

(12) H Le Chatelier, "Experimental Researches

in the Constitution of Hydraulic Cements,"

translation by J L Mack, McGraw-Hill

Book Co., Inc., New York, N Y (1905)

(13) G L Kalousek, C H Jumper, and J J

Tregoning, "Composition and Physical

Properties of Aqueous Extracts From Port-

land Cement Clinker Pastes Containing

Added Materials," Journal of Research,

Nat Bureau of Standards, Vol 30, p 215

(1943)

(14) P Schiapper, "Effect of Water on Portland

Cement," Symposium on the Chemistry of Cement, p 270, Stockholm (1938) (15) M I Strelkov, "Changes in the True State

of the Liquid Phase During Hardening of Cements and the Mechanism of Their Hardening," Reports of Symposium on the Chemistry of Cements, State Publication

of Literature on Structural Material, Mos- cow (1956)

(16) W Lerch, "The Influence of Gypsum on the Hydration and Properties of Portland Cement Pastes," Proceedings, Am Soc Testing Mats., Vol 46, p 1252 (1946) (17) W A Weyl, "A New Approach to Surface Chemistry and Heterogeneous Catalysis," BuLletin No 57, Mineral Industries Experi- ment Station, Pennsylvania State College, State College, Pa

(18) J W Jeffery, "Practical Implications of Fundamental Research in Cement Hydra- tion," Chemistry and Industry, p 1756 (1955)

(19) J D C McConneU, "The Hydration of Larnite (B-CasSiO4) and Bredigite ( a t CasSiO4) and the Properties of the Result- ing Gelatinous Mineral Plombierite,"

Mineralogical Magazine, Vol 30, p 672 (1955)

(20) W C Hansen, "Aeration Cause of False Set in Portland Cement," Proceedings, Am

SOc for Testing Mats Vol 38, p 1044 (1958); W C Hansen and E E Pressler,

"Solubility of Ca(OH)~ and CaSO4-2HsO

in Dilute Alkali Solutions," Induar/al and

Engineering Chemistry, Am Chemical Soc., Vol 39, p 1280 (1947)

(21) L S Brown and M A Swayze, "Detecting Free Lime and Magnesia in Portland Ce- ment," Rock Produas, Vol 41, No 6, p 65 (1938)

(se:l) Paul S Roller, "The Setting of Portland Cement Chemical Reactions of Seasoning, Reversion and Restoration," Industrial and Engineering Chemistry, Am Chemical SOc., Vol 26, p 1077 (1934)

(23) R I Razouk and R Sh Mikhail, "The Hydration of Magnesium Oxide From the Vapor Phase," Journal of Physical Chem- istry, Am Chemical SOc., Vol 62, p 920 (1958)

(24) T J Gray, et al, "The Defect Solid State," Interscience Publishers, New York, N Y (1957)

(25) H H Steinour, "Rate of Sedimentation

I Nonflocculated Suspensions of Uniform Spheres; II Suspensions of Uniform Size Angular Particles; HI Concentrated Floc-

DISCUSSION OF ACTIONS OF

culated Suspension of Powders," Industrial

Soc., Vol 36, pp 618, 840, 901 (1944)

(26) C M Midgley, "The Crystal Structure of

Dicalcium Silicate," Aaa Crystallographica,

Vol 5, Part 3, p 307 (1952)

(27) Stephen Brunauer, D L Kanto, and L E

Copeland, "The Stoichiometry of the Hy-

dration of Beta-Dicalcium Silicate and

Tricaldum Silicate at Room Temperature,"

(1958)

(28) L Heller, "The Structure of Dicalcium

Silicate a-Hydrate," A aa Crystallogra phi~a,

Vol 5, Part 6, p 724 (1952)

(29) Helen D Megaw, "The Structure of

Mwillite, Ca,(SiO3OH) 2- 2H20," A aa Crys-

(80) L R~ Forbrich, "The Effects of Various

Reagents on the Heat Liberation Charac-

teristics of Portland Cement," Proceedings,

Am Concrete Inst., Voi 37, p 161 (1940)

(31.) H L Kennedy, "Portland Cement

Effects of Catalysts and Dispersions," In-

Chemical Sot., Vol 28, p 963 (1936)

(32) F M Ernsberger and W G France, "Port- land Cement Dispersion by Adsorption of Calcium Liguosulfonate," Industrial and

Vol 37, p 598 (1945)

(33) Lunus Paufing, "The Nature of the Chemi- cal Bond," Cornell University Press, Ithaca, N Y., p 308 (1948)

(34) E P Flint and L S Wells, "Analogy of Hydrated Calcium Silica Aluminates and Hexacalcium Aluminate to Hydrated Cal- cium Suifoaluminates," Journal o] Research,

Nat Bureau of Standards, Vol 33, p 471 (1944)

(85) L H Tuthill and W A Cordon, "Proper- ties and Uses of Initially Retarded Con-

crete," Journal, Am Concrete Inst., Vol

27, p 273 (1956)

(36) O A Lukyanova, E E Segalova, and P

A Rehbinder, "Effect of Hydrophilic Plasticizing Admixtures on the Properties

of Concentrated Cement Suspensions,"

Moscow (1957) (Translation from Slavic Languages Assn Translation Center.)

DISCUSSION

shows, the past experimental work leaves

much uncertainty regarding the chemical

and physical events t h a t take place in

portland cement paste before it hardens

There are, of course, good reasons for

this Portland cement is a v e r y complex

material, and the paste is a thick suspen-

sion ill adapted to the s t u d y of what goes

on inside Thus the evidence t h a t we now

have is in large degree indirect or in-

complete

To present a detailed picture on the

basis of present knowledge involves,

therefore, considerable speculation Be-

fore we can eliminate the speculative ele-

ments we shall have to do much more

research I know from discussion with the

author t h a t he is fully as convinced as I

am of the great inadequacy of present in-

formation

1 Assistant to Director of Research, Portland

Cement Assn., Research and Development Lab-

oratories, Skokie, Ill

As is natural in so speculative an area, the author shows that different investiga- tors arrived at different interpretations

He accepts a part, rejects a part, and builds his own conception of things T h e foundation of his edifice is the assump- tion of "solid state reactions."

This is a bold move Although, as he shows, some others have also assumed

into the detail t h a t he has or been so articulate about diffusion into crystals as

a major feature of the reaction mech- anisms

I n thus presenting something new in

so controversial a field, I am sure t h a t the author d~d not expect, nor wish, to avoid critical comment Following his own pattern, I am inclined to accept a part and question a part of what he has offered

Retardation by Gypsum:

W h a t I can accept wholeheartediy is the prominant role assigned to adsorption I

26 SYMPOSIUM ON ADMIXTURES IN CONCRETE

think of this as formation of surface

layers or coatings (held by either physical

o~'~chemical adsorption and not limited

io monolayers), which coatings act to

hinder further reaction of the underlying

solid However, I think of these coatings

as forming by precipitation onto the re-

acting solids M y views are expressed in

a paper published in 1958 (1) 2 Though

m y ideas of mechanism differ from those

6f the author, I agree with his conclusion

that in retardation b y gypsum or other

form of calcium sulfate "some combina-

tion of SO~ with 3CaO-AI20, prevents

flash set." This product is called a cal-

cium sulfoaluminate Experiments have

shown that a large proportion of the

gypsum retarder reacts within the first

moments of contact with w~tter, forming

apparently such a product (2,3) This

sulfoaluminate apparently coats the sur-

faces of the tricalcium aluminate, thus

preventing ready access to the solution,

for the reaction with sulfate drops ab-

ruptly to a very low rate The paste re-

mains plastic, showing that flash set has

been avoided

The coating is not wholly protective, for

the tricalcium aluminate continues to re-

act slowly, but the reaction simply forms

more sulfoaluminate which can continu-

ally repair the coating as long as any

calcium sulfate remains However, exper-

iment shows that the sulfate is eventually

consumed There then occurs in some

cements the "delayed rapid reaction"

that Lerch (reference (10) of the paper)

demonstrated by means of the heat liber-

ation This is apparently a rapid reaction

of the tricalcium aluminate that can

occur when no more sulfoaluminate can

form and calcium aluminate hydrate

starts forming.instead Formation of this

new p r o d u c t ' a p p a r e n t l y breaks up the

coating, thus allowing the reaction to ac-

celerate A point of special interest is

of the hydration of tricalcium silicate, for experiments with pastes of this com- pound show that they set in about the same time as normal cement pastes

Of course, when too little gypsum is used, the control exercised over the re- action of the tricalcium aluminate m a y cease at a relatively early stage, with relatively early onset of t h e "delayed rapid reaction." If the cement has not already set, this reaction of the aluminate can evidently produce set much as in un- retarded cement The nature of the re- action causing set is thus dependent on the amount of gypsum used However, cements that contain what Lerch defined

as the optimum gypsum content should apparently all set because of hydration

of the tricalcium silicate the aluminate reaction being sufficiently controlled

Slowness of Solid State Reactions:

Returning now to the author's theory

of solid state reactions, I regret t h a t a part of it is very difficult for me to accept

He assumes that ions can diffuse with great rapidity into the clinker compounds

rough calculations indicate that diffusion through m a n y molecular layers would be necessary to account b y this mechanism for the amount of reaction that he ob- tained in 1 min This rapid diffusion into the solids is so basic an element of his theory that I regret to dispute it, but to

me it seems very unlikely

Our laboratory library has several

I)ISCUSSION OF ACTIONS OF CALCIIYM SULFATE 27 books and journal papers that review

solid state reactions Examination of

these has confirmed m y previous,impres-

sion that such reactions have nearly

always been studied at elevated temper-

atures for the reason that they do not

generally occur with sufficient or even

measurable rapidity at room tempera-

ture The reference here is to solid re-

actions that require diffusion into the

solid, as distinguished from reactions

such as dehydrations that simply remove

a part of the crystal lattice and can open

up channels for further removal I ex-

empt, also, reactions such as those of

zeolites, where the crystal lattices are so

open-structured t h a t ions can move

through the channels as they might along

an exterior surface Though the atoms in

the cement compounds m a y not all be in

the closest packing, the crystals certainly

contain no such channels as in the zeo-

lites The densities of the cement com-

pounds are relatively high, and diffusion

of such ions as H~O + and O H - into the

crystals would apparently have to occur

b y displacements such as occur in those

solid state reactions t h a t proceed very

slowly or not at all at room temperature

The author's discussion is in terms of

diffusion into these original solids, but I

find that the diffusion that is generally

discussed in solid state reaction theory is

that through the layer of reaction prod-

uct that builds up Indeed, it is hard for

me to see how there could be much dif-

fusion into the cement compounds since

the products are of greater volume, and

there would apparently be a breaking up

at the surface If diffusion into the orig-

inal solids does not have to occur and if

the products (other than Ca(Oil)2) are

gels and thus relatively permeable, then

I can see more possibility of some kind

author has assumed diffusion into the

original solids Hence m y discussion re-

lates to diffusion of this kind

Since solid state reactions are studied

in the absence of a liquid phase, the per- tinent literature relates mainly to reac- tions between solids M y examination disclosed one so-called reaction or "start- ing" temperature as low as 160 C Other reactions were indicated to proceed be- tween 200 and 300 C and others between

300 and 400 C These seemed to be re- garded as the lower ranges of tempera- tures for solid state reactions involving interaction of inorganic salts and oxides

T a m m a n n , an early experimenter in this field, reported an approximate rela- tionship between reaction temperature and melting point of the solid Welsh (4) reporting in 1955 defined the " T a m m a n n

the minimum temperature at which the solid m a y be expected to enter into a solid state reaction of appreciable rate."

He said, "Such characteristic tempera- tures are only a general guide to solid reactivity, but as such they are useful." For reaction involving lattice diffusion

he gives the T a m m a n n temperature as one half the melting point on the abso- lute scale For appreciable "surface mo- bility of lattice units," the corresponding fraction is three tenths of melting point Since the major compounds in cement clinker have melting or decomposition points ranging from about 1400 C to above 2000 C, the reaction temperature

on either basis would be far above room temperature

Hedvall, another extensive investiga- tor in this field, found that for a number

of reactions involving calcium oxide the reaction temperatures were all somewhat above 500 C (4)

Viewed against this background, solid state reactions involving diffusion into the crystal do not appear to me to pro- vide a sufficiently plausible explanation

of the rapid formation of reaction prod- uct when portland cement first comes into contact with water

28 S ~ P o s t ~ oN ADM~xTm~s tN COgCRZTZ

Volume Considerations:

Support for this viewpoint can also be

drawn from the fact that the hydration

products occupy much greater volume

than the anhydrous solids Even if the

calcium hydroxide that is split off is

thought of as dissolving and going else-

where, there is still a large volume in-

crease This is true even for tricalcium

silicate, owing to the pore volume that

Powers (S) has found to be an essential

part of the hydration product, or cement

gel Thus, if solid state reaction is the

correct mechanism throughout, the orig-

inal particle must obviously expand

greatly and can only do so by pressing

out against its surroundings A similar

action evidently does occur in the case

of delayed hydration of calcium and mag-

nesium oxides, as the author has noted

These, however, are disruptive and dis-

integrative reactions whereas the normal

hydration of portland cement is not

Hydration by dissolution and precipi-

tation, which is the usual mechanism of

low-temperature reactions in the pres-

ence of a liquid phase, does not involve

this problem Of course, as reaction pro-

ceeds and the cement particles become

enveloped with reaction product, reac-

tion by solution and precipitation must

become more involved Presumably, sur-

face forces are then felt at all times

Solution may then take on a somewhat

different meaning while still involving

the loosening of ions from their positions

in the crystal, with passage into a more

fluid film before being reoriented and de-

posited as part of a hydrate molecule

Some hydrate could be deposited in the

space made available by the dissolution,

and some would evidently have to form

by transport of ions out through narrow

channels until spaces affording sufficient

room for deposition of product were

reached Powers (6) has given consider-

able thought to matters of this kind

Here, we are concerned with the early

reactions before this problem becomes acute

However, before leaving the subject,

it is of interest to refer to the observa- tions of McConnell (reference (19) of the author's paper) McConnell found natur- ally occurring nodules of dicalcium sili-

sheaths of considerable thickness In this case, leaching of the original crystals was a prominent feature of the mech- anism for there were left only 6 silicon and 5 calcium atoms in unit volume that originally contained 8 silicon and 16 cal-

cleared for the hydrate water

Lack of Direct Data:

The author rejects the solution and precipitation mechanism because he thinks it could not operate fast enough

to give as much product as was formed during the first minute It might perhaps

be argued that the experiments on which

he bases this conclusion involved reaction with calcium chloride instead of the usual reaction with gypsum; however, I do not raise this point because I believe, as he does, that the usual reaction rate is com- parable

As I understand it, the autho is not arguing, as a general truth, that reactions

of solution and precipitation cannot occur so fast as the reaction that did occur This would be a very difficult posi- tion to maintain, for it is common knowl- edge that solution and precipitation can progress with great rapidity What he finds impossible to believe is that the re- actions of compounds containing silica and alumina could occur so fast "when only traces of these oxides can be found

in the liquid phase."

Study of this qualification indicates that the assumed impossibility of the reaction reduces to assumed impossibil- ity of the very high rate of transfer or diffusion of ions through the solution

DISCUSSION OF ACTIONS OF CALCIUM SULFATE 29

which would be necessary if the concen-

trations during reaction stay so low

Calculations could no doubt be made

that would establish the possibilities or

impossibilities more definitely, but I am

much more inclined to dispute the prem-

ise than the logic that is based on it

That is, I do not believe that the con-

centrations did remain, during the ini-

tial minute of reaction, at such trace

amounts

The only numerical data that the

paper presents on this point are for ex-

tracts taken at 7 min and 2 hr But 7

min is too late a time to furnish a reli-

able indication of the situation during

the first minute, which is the period for

which the amount of reaction was deter-

mined The author recognized this, for

he stated that similar data are obtained

for shorter reaction periods I do not

doubt this, but I do doubt that data

that support his position have been ob-

tained within the initial 1-min period,

which is what is required

In an earlier report (reference (3) of

the paper, pp 318-321) the author used

the 1-min data of his Table I as indicat-

ing low silica and alumina concentrations

because rather good electrostatic bal-

ances of positive and negative ions are

obtained without counting these ions This indication, though more subject to error than a direct determination, war- rants attention, but it too does not take

us within the first minute In other words, it does not show how high the concentrations may have risen during the period of solution and precipitation with which we are concerned

Data for More Dilute Suspensions Support the Solution Theory:

I am not aware of any extraction data having been obtained short of 1 min, but extracts taken at longer time intervals from more dilute suspensions give indica- tions of what happens earlier in dense paste such as cement paste Whatever the concentration of particles in the ini- tial suspension, the concentration in solution builds up toward a peak by striving toward the solubility-concentra- tion of the initial solid, and then as pre- cipitation occurs it drops toward the lower solubility-concentration of the product This course is indicated by the accompanying Fig 2 Decreasing the concentration of solid in the suspension naturally lengthens the time that would

be required to reach saturation if pre- cipitation did not occur Obviously, too,

it lengthens the time required to reach the same height of peak when precip- itation does develop Owing to this stretched-out time scale (which may favor precipitation at less supersatura- tion) the peak may not go so high, but

it is probable that the decline from the peak will also be stretched out because

of increase of distance between precipi- tation points That is, precipitation prob- ably occurs largely on the original par- ticles and thus occurs less rapidly when these are farther apart On the whole, therefore, it appears that for the more dilute suspensions there would be a rather general stretching of t h e time scale of Fig 2 Thus, it is indicated that

L 5 1 6 8 ) 5 2 3 6 ) 5 5 8 6 ) 5 6 3 4 ) 5 7 4 4 ) 6 2 8 4

I 7 6 2 4 6 1 2 5

In the author's experiments with cal- cium chloride, he used 200 g of clinker

to 150 ml of solution, or 1333 g of clinker per liter In the work of Kalousek, Jumper, and Tregoning (reference (13) of the paper), for which the author reports the 7-min and 2-hr data, a water-cement

CaO in Solution, g per liter

F I o 3. Solution Derived from/~ C2S and CsS (20 g per liter) (7)

0 6

high concentrations representing super-

saturation relative to the products will

be found at later times when the dilution

of the initial suspension is increased

Making tests at a fixed period on sus-

pensions of progressively increased dilu-

tion would then be comparable in effect

to moving backward on the time scale of

Fig 2 I n any case, if concentrations are

permanently low after 1 min in dense

paste but are found to be relatively high

ratio of 0.35 was used, or 2860 g of clinker

or cement per liter I t is therefore of in- terest to examine data of Flint and Wells (7) which were obtained b y shaking only

20 g of cement with 1 liter of water These data are shown in the accompany- ing Table IV At 1 min the concentra- tions of Al2Oa and SiO2 m a y still appear small, but they are much larger than the

3 mg and less shown b y the 7-min data

of Kalousek and his associates I t is

DISCUSSION OF ACTIONS OF CALCIIYM SULFATE 31

apparent also that the values diminish

with time, thus suggesting that at 1 min

the values were already on the wane It

is to be noted, however, that the CaO

values are increasing with time This is

owing to hydrolysis with production of

calcium hydroxide which is more soluble

than the other products Thus, although

the other hydrates are already precipi-

tating, the calcium hydroxide concentra-

tion has not yet reached saturation The

increase in calcium hydroxide concentra-

tion depresses the solubilities of the sili-

cate and aluminate hydrates and causes

their concentrations to drop more rapidly

than would otherwise be the case In the

7-min Kalousek data, the CaO values

were high, which was sufficient reason for

the low values of SiO2 and A1203 Before

these high CaO values were attained, the

SiO2 and A1203 values would have been

higher because of the higher solubilities,

even had there been no supersaturation

Valuable information is also provided

by Fig 3 which shows data obtained by

Flint and Wells (7) upon shaking trical-

cium silicate and beta-dicalcium silicate

with water, using the same proportions

as for cement, namely, 20 g per liter

Here the SiO2 concentration is plotted

against the CaO concentration, and the

points are labeled to show the times at

which the extracts were taken Dashed

tie lines connect the curves representing

the analyses of the extracts with a lower

curve that shows the equilibrium concen-

trations attained after precipitation oc-

curred in the extracts

In the case of the slowly reacting di-

calcium silicate, the relative positions of

the 1-min and 5-min points indicates

that the SiO~ concentration probably did

not go much higher than was found at

these times, namely, about 85 mg per

liter But in the case of the tricalcium

silicate it is evident that the SiO2 con-

centration may have gone considerably

higher than the 64 mg per liter found at

3 min It is the rapid increase of the cal- cium hydroxide concentration that has forced this value down below that found for the dicaicium silicate at 1 min The

64 mg per liter is, however, a very high value as compared with the value of less than 1 mg per liter shown by Kalousek's 7-min data reported by the author Forsen (reference (11) of the paper, pp 298-363) reported work in which only 0.2 g of tricalcium silicate was shaken with 1 liter of water Even this very weak suspension gave a 1-min extract containing 18.4 mg SiO2 per liter This

is 34.8 per cent of the total SiO, in the sample Although this percentage is sur- prisingly large, comparison with data for other periods indicates that if filtration was adequate it is perhaps a fairly reli- able result Under the prevailing condi- tions, the precipitation was small, for the total amount of tricalcium silicate that reacted 'was calculated as only 5.4 per cent more, that is, 40.2 per cent Here then is evidence that exceedingly rapid solution of tricalcium silicate may occur when blocking by the product is reduced

to a minimum

Forsen also reported that 0.2 g of tri- calcium aluminate shaken with a liter of water for 1 min gave an A1203 concen- tration of 23.7 mg per liter, which is 31.2 per cent of the total Al~O3 I t is a very high concentration as compared with that in the 7-min extract reported by the author

These various data lead me to believe that the early reactions occurring in ce- ment paste at ordinary temperature do not depart from the usual rule but occur mainly by solution and precipitation It

is natural that in a cement paste these initial reactions should largely cease within a very short time since 1 min is a reasonably adequate mixing period for concrete, and a proper mixing period must be long enough for establishment

of the conditions of retardation With

32 Sx~tPOSim~ oN ADMIXTtrRES IN CONCREr~

cessation of the reaction, the concentra-

tions of silica and alumina quickly ap-

proach the very low values that corre-

spond to the solubilities of the products

The data indicate that at earlier mo-

ments the concentrations are much

higher T h e y m a y indeed be considerably

higher than any of the values reported,

since there is no reason to believe that

these represent the peak amounts at-

tainable in the paste

The reaction of tricalcium silicate is

peculiar in that the initial rapid reaction

is so quickly reduced to a low rate that

it does not cause setting and yet later on

the reaction again accelerates, is not

quickly damped, and does cause setting

This course is followed even in pure tri-

calcium silicate paste, and the changes

in reaction rate are shown b y data on

heat liberation (reference (30) of the

paper, pp 161-184) This self-regulating

effect of tricalcium silicate has received

no satisfactory explanation One m a y

assume that the initial reaction is cut

short b y the blanketing effect of hydra-

tion product, but the reason why the

effect should later be overcome is not

clear Some possibilities suggest them-

selves, but only further research can

provide a decisive answer

Importance of Hydroxyl Groups:

I am glad now to turn from these con-

siderations to an area in which the author

and I are in rather complete agreement

The author greatly advanced our under-

standing of the action of the majority of

organic retarders when, in 1952 (refer-

ence (3) of the paper, pp 598-627), he

reported that most retarders used in oil-

well cements fall in three main classes

and that all three classes consist of com-

pounds that contain the atom group

H C O H As he reports in the present

paper, I presented additional data which

indicated that the active factor is simply

the O H group, for OH groups attached

to the benzene ring appear to be effective

in this way also (reference (a) of the paper, pp 627-631) In tests of chemical additions to pastes of beta-dicalcium sili- cate I had found an apparent relation- ship between degree of retardation and number of un-ionized or undissociated

O H groups I t appeared also that OH groups in some inorganic compounds are similarly effective if they do not dissoci- ate or ionize off The most plausible mechanism appeared to be t h a t of ad- sorption through formation of a hydrogen bond From the author's study of cement retarders, it seems clear that what was found for dicalcium silicate can be ex- tended to cements Indeed, in cements with good contents of gypsum in which setting is due to hydration of the trical- cium silicate, it is logical to assume that

in order to obtain further retardation it

is the reaction of the silicate that must

be slowed up An agent that is adsorbed

on dicalcium silicate would presumably

be adsorbed on tricalcium silicate also There is also good reason to believe that adsorption through hydrogen bonding occurs on tricalcium aluminate A com- mon form of hydrogen bonding is attach- ment to oxygen atoms, and all of the major clinker compounds are combina- tions of oxides As the author has pointed out, the lignosulfonates contain OH groups Russian investigators have made extensive studies with the lignosulfonates and report extensive adsorption on tri- calcium aluminate (8)

The molecules of chemicals that con- tain OH groups tend to form hydrogen bonds with one another Thus com- pounds with more than one O H group

m a y perhaps form particularly stable adsorbed layers owing to the ability to form cross linkages Perhaps this is one reason for the apparent relationship be- tween number of OH groups and effi- ciency of retardation As the author pointed out in his 1952 paper, the fact

I)ISCUSSION OF ACTIONS OF CALCIUM SULFATE 33

that a compound contains an OH group

does not insure that it wilt be a retarder,

for there m a y be opposing influences

However, a compound containing several

OH groups appears generally to be a

strong retarder Ordinary sugar (sucrose)

is reputed to be a very destructive re-

tarder I t sometimes gives a very quick

set owing apparently to the amount of

reaction that can sometimes occur before

the retardation is established, but fairly

small amounts can seriously affect early

strength development This compound,

Cx2H220n, contains eight O H groups

MR W C HANSEN (author). I am

pleased that the case for the solution-

precipitation hypothesis has been pre-

sented I cannot, of course, present data

to refute Mr Steinour's arguments be-

cause the same data are available to

both of us In other words, as both agree,

there is need for experimental work in

this field

• This position by Mr Steinour is some-

what surprising because he and T C

Powers (9) published a paper in which

they seemed to accept the previously

published hypothesis that grains of opal

encased in a hardened cement paste

could chemisorb alkali and hydroxyl

ions from the liquid phase of the hard-

ened paste and form solid alkali silicates

I t seems to me that the mechanism sug-

gested for the reactions between cement

minerals and ions from the liquid phase

of cement pastes is similar to that postu-

lated for the reaction of opal and alkalies

M t e r m y paper was written, I obtained

a translation of a paper b y Funk (10)

who reacted ~ 2CaO.SiO~ with water

vapor at 100 C and obtained a product

that had an X - r a y diffraction pattern

similar to that of the tobermorite-like

product obtained when ~ 2CaO-SiO~ is

hydrated in water He states:

"Le Chatelier's hypothesis is generally ac-

cepted as correct, namely, that 3CaO-SiO2

and 2CaO-SiO2 go into solution during hy-

drolysis and that calcium silicate hydrate is precipitated out of solution due to its low solubility Formation of the tobermorite-like calcium silicate hydrate phase through ac- tion of water vapor on~ 2CaO-SiO2 at 100 C demonstrates that some other mechanism may be at work during the reaction of Hg) with/~ 2CaO SiO2 "

Chemists long ago recognized that re- actions occurred at the surfaces of solids

In some cases, they were certain that these were chemical reactions and used the word chemisorption for such reac- tions When they were in doubt about them being chemical reactions, they used the term adsorption However, they did limit these reactions, at least b y defini- tion, to reactions of a solid with ions and molecules from a liquid or a gaseous phase T h a t is, the definition of adsorp- tion did not include a n y reaction of one solid with another as adsorption The solution-precipitation theory appears to include a reaction between two solids under the heading of adsorption T h a t is,

it seems to visualize, for example, the precipitation of a hydrated calcium sili- cate on the surface of a crystal of 3CaO SiO~ as adsorption I cannot visualize this type of a process as one that should

be classed as either adsorption or chemi- sorption I t seems that a precipitation could occur in which the surface of the 3CaO-SiO~ would be covered with the hydrated silicate without any interaction between the two solids The precipitate could interfere with the further reaction

of the 3CaO-SiO~ and water b y a more

or less blanketing effect

A number of investigators have sug- gested that the hydrated calcium silicates formed by the reactions of 2CaO-Si02 and 3CaO.SiO~ contained adsorbed Ca(OH)z I t was never clear whether this was visualized as a case of one solid adsorbing another solid or whether cal- cium and hydroxyl ions were adsorbed from solution by the solid hydrated cal-

34 SYm, osro~ ON ADMIXTURES IN CONCRETE

cium silicate Recently, Brunauer, Kan-

and from studies of the work of others,

concluded that the adsorption of Ca(OH) 2

on the surface of tobermorite must not

be appreciable The results of their

studies also indicate that some Ca(OH)2

would exist in hardened cement pastes

as amorphous or poorly crystallized

Ca(OH)2 This, together with the evi-

dence that the other reaction products are

also either amorphous or poorly crystal-

lized, does not support the solution-

precipitation process

One can also visualize the case in which

a hydrated silicate could crystallize on

the surface of a crystal of 3CaO SiO2 by

using the 3CaO.SiO~ as a nucleus on

which to crystallize This, it seems,

should not be called adsorption How-

ever, if it did occur, there is the problem

of the mechanism whereby the 3CaO-

SiO2 that acted as the nucleus reacts

with water for its final hydration

Mr Steinour argues against what I

have classed as solid state reactions oc-

curring at atmospheric temperatures

However, he seems to have accepted the

idea that opal and sodium hydroxide can

react at such temperatures without the

opal going into solution He likewise

accepts the reaction of MgO with water

vapor at -these temperatures I see no

reason for doubting the possibility that

minerals, such as 3CaO- SiO2 and 3CaO

Al2Oa, can undergo a similar type of

reaction with water at atmospheric

temperatures

Mr Steinour cannot visualize these

solid state reactions occurring without

causing expansions Our picture of the

structure of hardened cement paste is not

based on direct evidence It is built up

primarily from sorption data These data

may give a fairly accurate measure of

porosity of the paste However, it does

not seem that they can possibly reveal

whether or not a crystal of 3CaO-SiO2

has reacted with water to give a cluster

of identical particles of a hydrated sili- cate in which the porosity of the cluster

is uniform throughout It seems that the same sorption results might be obtained

if the porosity of that cluster varied widely from the center to outside and also if the amount of water combined in the individual particles varied widely The picture as revealed by these data is

an average picture and, as such, cannot,

in my opinion, be used to furnish data for calculating the volume change of any given particle of cement as it reacts in a cement paste It is necessary to build these pictures to help in research but they should not be accepted as estab- lished facts

Mr Steinour uses data in his Table IV and Fig 3, taken from work by Flint and Wells, to support the solution theory In our laboratory we have made many ex- traction tests on pastes with water-ce- ment ratios of about 0.50 to 0.75 The compositions of these extracts show that generally the liquid phase of a cement paste is slightly supersaturated with cal- cium hydroxide at the 1-min period Therefore, it seems that the liquid phase

of the cement pastes in normal mortars and concretes must be saturated with respect to calcium hydroxide almost im- mediately when cement and water come

in contact with each other If this is true, then attempts to study extracts at periods of less than 1 min might not yield information different from that ob- tained from extracts taken approximately

1 min after mixing

The data from Flint and Wells used

by Mr Steinour were obtained from so- lutions that were not saturated with calcium hydroxide for periods of several hours These data show that the concen- trations of SiO2 and A1203 are each about 0.002 g per liter when the solutions are saturated with calcium hydroxide These values compare with those reported by